ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/OpenCL/OpenCL.xs
(Generate patch)

Comparing OpenCL/OpenCL.xs (file contents):
Revision 1.37 by root, Sat Apr 21 18:49:21 2012 UTC vs.
Revision 1.81 by root, Sat May 5 17:09:19 2012 UTC

1#include <stddef.h>
2
1#include "EXTERN.h" 3#include "EXTERN.h"
2#include "perl.h" 4#include "perl.h"
3#include "XSUB.h" 5#include "XSUB.h"
6
7#include "ecb.h"//D
8
9#define X_STACKSIZE sizeof (void *) * 512 * 1024 // 2-4mb should be enough, really
10#include "xthread.h"
11#include "schmorp.h"
4 12
5#ifdef I_DLFCN 13#ifdef I_DLFCN
6 #include <dlfcn.h> 14 #include <dlfcn.h>
7#endif 15#endif
8 16
11// this. This breaks 100% of the opencl 1.1 apps, for what reason? 19// this. This breaks 100% of the opencl 1.1 apps, for what reason?
12// after all, the functions are deprecated, not removed. 20// after all, the functions are deprecated, not removed.
13// in addition, you cannot test for this in any future-proof way. 21// in addition, you cannot test for this in any future-proof way.
14// each time a new opencl version comes out, you need to make a new 22// each time a new opencl version comes out, you need to make a new
15// release. 23// release.
16#define CL_USE_DEPRECATED_OPENCL_1_1_APIS
17#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */ 24#define CL_USE_DEPRECATED_OPENCL_1_2_APIS /* just guessing, you stupid idiots */
18 25
26#ifndef PREFER_1_1
27 #define PREFER_1_1 1
28#endif
29
30#if PREFER_1_1
31 #define CL_USE_DEPRECATED_OPENCL_1_1_APIS
32#endif
33
19#ifdef __APPLE__ 34#ifdef __APPLE__
20 #include <OpenCL/opencl.h> 35 #define CLHDR(name) <OpenCL/name>
21#else 36#else
22 #include <CL/opencl.h> 37 #define CLHDR(name) <CL/name>
23#endif 38#endif
39
40#include CLHDR(opencl.h)
41
42#if 0
43#ifndef CL_VERSION_1_2
44 #include CLHDR(cl_d3d9.h)
45#endif
46#endif
47
48#if _WIN32
49 #include CLHDR(cl_d3d10.h)
50 #if CL_VERSION_1_2
51 #include CLHDR<cl_d3d11.h>
52 #endif
53 #include CLHDR<cl_dx9_media_sharing.h.h>
54#endif
55
56#ifndef CL_VERSION_1_2
57 #undef PREFER_1_1
58 #define PREFER_1_1 1
59#endif
60
61// make sure all constants we might use are actually defined
62#include "default.h"
24 63
25typedef cl_platform_id OpenCL__Platform; 64typedef cl_platform_id OpenCL__Platform;
26typedef cl_device_id OpenCL__Device; 65typedef cl_device_id OpenCL__Device;
66typedef cl_device_id OpenCL__SubDevice;
27typedef cl_context OpenCL__Context; 67typedef cl_context OpenCL__Context;
28typedef cl_command_queue OpenCL__Queue; 68typedef cl_command_queue OpenCL__Queue;
29typedef cl_mem OpenCL__Memory; 69typedef cl_mem OpenCL__Memory;
30typedef cl_mem OpenCL__Buffer; 70typedef cl_mem OpenCL__Buffer;
31typedef cl_mem OpenCL__BufferObj; 71typedef cl_mem OpenCL__BufferObj;
32typedef cl_mem OpenCL__Image; 72typedef cl_mem OpenCL__Image;
33typedef cl_mem OpenCL__Image2D;
34typedef cl_mem OpenCL__Image3D;
35typedef cl_mem OpenCL__Memory_ornull; 73typedef cl_mem OpenCL__Memory_ornull;
36typedef cl_mem OpenCL__Buffer_ornull; 74typedef cl_mem OpenCL__Buffer_ornull;
37typedef cl_mem OpenCL__Image_ornull; 75typedef cl_mem OpenCL__Image_ornull;
38typedef cl_mem OpenCL__Image2D_ornull;
39typedef cl_mem OpenCL__Image3D_ornull;
40typedef cl_sampler OpenCL__Sampler; 76typedef cl_sampler OpenCL__Sampler;
41typedef cl_program OpenCL__Program; 77typedef cl_program OpenCL__Program;
42typedef cl_kernel OpenCL__Kernel; 78typedef cl_kernel OpenCL__Kernel;
43typedef cl_event OpenCL__Event; 79typedef cl_event OpenCL__Event;
44typedef cl_event OpenCL__UserEvent; 80typedef cl_event OpenCL__UserEvent;
45 81
46typedef SV *FUTURE; 82typedef struct mapped * OpenCL__Mapped;
83
84static HV
85 *stash_platform,
86 *stash_device,
87 *stash_subdevice,
88 *stash_context,
89 *stash_queue,
90 *stash_program,
91 *stash_kernel,
92 *stash_sampler,
93 *stash_event,
94 *stash_userevent,
95 *stash_memory,
96 *stash_buffer,
97 *stash_bufferobj,
98 *stash_image,
99 *stash_image1d,
100 *stash_image1darray,
101 *stash_image1dbuffer,
102 *stash_image2d,
103 *stash_image2darray,
104 *stash_image3d,
105 *stash_mapped,
106 *stash_mappedbuffer,
107 *stash_mappedimage;
47 108
48/*****************************************************************************/ 109/*****************************************************************************/
49 110
50// name must include a leading underscore 111// name must include a leading underscore
51// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module 112// all of this horrors would be unneceesary if somebody wrote a proper OpenGL module
91} 152}
92 153
93/*****************************************************************************/ 154/*****************************************************************************/
94 155
95/* up to two temporary buffers */ 156/* up to two temporary buffers */
96static void * 157static void * ecb_noinline
97tmpbuf (size_t size) 158tmpbuf (size_t size)
98{ 159{
99 enum { buffers = 3 }; 160 enum { buffers = 4 };
100 static int idx; 161 static int idx;
101 static void *buf [buffers]; 162 static void *buf [buffers];
102 static size_t len [buffers]; 163 static size_t len [buffers];
103 164
104 idx = (idx + 1) % buffers; 165 idx = (idx + 1) % buffers;
111 } 172 }
112 173
113 return buf [idx]; 174 return buf [idx];
114} 175}
115 176
177static const char * ecb_noinline
178cv_get_name (CV *cv)
179{
180 static char fullname [256];
181
182 GV *gv = CvGV (cv); // gv better be nonzero
183
184 HV *stash = GvSTASH (gv);
185 const char *hvname = HvNAME_get (stash); // stash also better be nonzero
186 const char *gvname = GvNAME (gv);
187
188 snprintf (fullname, sizeof (fullname), "%s::%s", hvname, gvname);
189 return fullname;
190}
191
116/*****************************************************************************/ 192/*****************************************************************************/
117 193
118typedef struct 194typedef struct
119{ 195{
120 IV iv; 196 IV iv;
121 const char *name; 197 const char *name;
122 #define const_iv(name) { (IV)CL_ ## name, # name }, 198 #define const_iv(name) { (IV)CL_ ## name, # name },
123} ivstr; 199} ivstr;
200
201typedef struct
202{
203 NV nv;
204 const char *name;
205 #define const_nv(name) { (NV)CL_ ## name, # name },
206} nvstr;
124 207
125static const char * 208static const char *
126iv2str (IV value, const ivstr *base, int count, const char *fallback) 209iv2str (IV value, const ivstr *base, int count, const char *fallback)
127{ 210{
128 int i; 211 int i;
177 if (res) \ 260 if (res) \
178 FAIL (name); 261 FAIL (name);
179 262
180/*****************************************************************************/ 263/*****************************************************************************/
181 264
182#define NEW_MORTAL_OBJ(class,ptr) sv_setref_pv (sv_newmortal (), class, ptr)
183#define XPUSH_NEW_OBJ(class,ptr) XPUSHs (NEW_MORTAL_OBJ (class, ptr))
184
185static void * 265static SV *
266new_clobj (HV *stash, IV id)
267{
268 return sv_2mortal (sv_bless (newRV_noinc (newSViv (id)), stash));
269}
270
271#define PUSH_CLOBJ(stash,id) PUSHs (new_clobj ((stash), (IV)(id)))
272#define XPUSH_CLOBJ(stash,id) XPUSHs (new_clobj ((stash), (IV)(id)))
273
274/* cl objects are either \$iv, or [$iv, ...] */
275/* they can be upgraded at runtime to the array form */
276static void * ecb_noinline
186SvPTROBJ (const char *func, const char *svname, SV *sv, const char *pkg) 277SvCLOBJ (CV *cv, const char *svname, SV *sv, const char *pkg)
187{ 278{
279 // sv_derived_from is quite slow :(
188 if (SvROK (sv) && sv_derived_from (sv, pkg)) 280 if (SvROK (sv) && sv_derived_from (sv, pkg))
189 return (void *)SvIV (SvRV (sv)); 281 return (void *)SvIV (SvRV (sv));
190 282
191 croak ("%s: %s is not of type %s", func, svname, pkg); 283 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, pkg);
284}
285
286// the "no-inherit" version of the above
287static void * ecb_noinline
288SvCLOBJ_ni (CV *cv, const char *svname, SV *sv, HV *stash)
289{
290 if (SvROK (sv) && SvSTASH (SvRV (sv)) == stash)
291 return (void *)SvIV (SvRV (sv));
292
293 croak ("%s: %s is not of type %s", cv_get_name (cv), svname, HvNAME (stash));
192} 294}
193 295
194/*****************************************************************************/ 296/*****************************************************************************/
195 297
196static cl_context_properties * 298static cl_context_properties * ecb_noinline
197SvCONTEXTPROPERTIES (const char *func, const char *svname, SV *sv, cl_context_properties *extra, int extracount) 299SvCONTEXTPROPERTIES (CV *cv, const char *svname, SV *sv, cl_context_properties *extra, int extracount)
198{ 300{
199 if (!sv || !SvOK (sv)) 301 if (!sv || !SvOK (sv))
200 if (extra) 302 if (extra)
201 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully 303 sv = sv_2mortal (newRV_noinc ((SV *)newAV ())); // slow, but rarely used hopefully
202 else 304 else
208 int i, len = av_len (av) + 1; 310 int i, len = av_len (av) + 1;
209 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1)); 311 cl_context_properties *p = tmpbuf (sizeof (cl_context_properties) * (len + extracount + 1));
210 cl_context_properties *l = p; 312 cl_context_properties *l = p;
211 313
212 if (len & 1) 314 if (len & 1)
213 croak ("%s: %s is not a property list (must be even number of elements)", func, svname); 315 croak ("%s: %s is not a property list (must contain an even number of elements)", cv_get_name (cv), svname);
214 316
215 while (extracount--) 317 while (extracount--)
216 *l++ = *extra++; 318 *l++ = *extra++;
217 319
218 for (i = 0; i < len; i += 2) 320 for (i = 0; i < len; i += 2)
221 SV *p_sv = *av_fetch (av, i + 1, 0); 323 SV *p_sv = *av_fetch (av, i + 1, 0);
222 cl_context_properties v = SvIV (p_sv); // code below can override 324 cl_context_properties v = SvIV (p_sv); // code below can override
223 325
224 switch (t) 326 switch (t)
225 { 327 {
328 case CL_CONTEXT_PLATFORM:
329 if (SvROK (p_sv))
330 v = (cl_context_properties)SvCLOBJ (cv, svname, p_sv, "OpenCL::Platform");
331 break;
332
226 case CL_GLX_DISPLAY_KHR: 333 case CL_GLX_DISPLAY_KHR:
227 if (!SvOK (p_sv)) 334 if (!SvOK (p_sv))
228 { 335 {
229 void *func = glsym ("_glXGetCurrentDisplay"); 336 void *func = glsym ("_glXGetCurrentDisplay");
230 if (func) 337 if (func)
253 *l = 0; 360 *l = 0;
254 361
255 return p; 362 return p;
256 } 363 }
257 364
258 croak ("%s: %s is not a property list (either undef or [type => value, ...])", func, svname); 365 croak ("%s: %s is not a property list (either undef or [type => value, ...])", cv_get_name (cv), svname);
366}
367
368// parse an array of CLOBJ into a void ** array in C - works only for CLOBJs whose representation
369// is a pointer (and only on well-behaved systems).
370static void * ecb_noinline
371object_list (CV *cv, int or_undef, const char *argname, SV *arg, const char *klass, cl_uint *rcount)
372{
373 if (!SvROK (arg) || SvTYPE (SvRV (arg)) != SVt_PVAV)
374 croak ("%s: '%s' parameter must be %sa reference to an array of %s objects",
375 cv_get_name (cv), argname, or_undef ? "undef or " : "", klass);
376
377 AV *av = (AV *)SvRV (arg);
378 void **list = 0;
379 cl_uint count = av_len (av) + 1;
380
381 if (count)
382 {
383 list = tmpbuf (sizeof (*list) * count);
384 int i;
385 for (i = 0; i < count; ++i)
386 list [i] = SvCLOBJ (cv, argname, *av_fetch (av, i, 1), klass);
387 }
388
389 if (!count && !or_undef)
390 croak ("%s: '%s' must contain at least one %s object",
391 cv_get_name (cv), argname, klass);
392
393 *rcount = count;
394 return (void *)list;
259} 395}
260 396
261/*****************************************************************************/ 397/*****************************************************************************/
398/* callback stuff */
399
400/* default context callback, log to stderr */
401static void CL_CALLBACK
402context_default_notify (const char *msg, const void *info, size_t cb, void *data)
403{
404 fprintf (stderr, "OpenCL Context Notify: %s\n", msg);
405}
406
407typedef struct
408{
409 int free_cb;
410 void (*push)(void *data1, void *data2, void *data3);
411} eq_vtbl;
412
413typedef struct eq_item
414{
415 struct eq_item *next;
416 eq_vtbl *vtbl;
417 SV *cb;
418 void *data1, *data2, *data3;
419} eq_item;
420
421static void (*eq_signal_func)(void *signal_arg, int value);
422static void *eq_signal_arg;
423static xmutex_t eq_lock = X_MUTEX_INIT;
424static eq_item *eq_head, *eq_tail;
425
426static void ecb_noinline
427eq_enq (eq_vtbl *vtbl, SV *cb, void *data1, void *data2, void *data3)
428{
429 eq_item *item = malloc (sizeof (eq_item));
430
431 item->next = 0;
432 item->vtbl = vtbl;
433 item->cb = cb;
434 item->data1 = data1;
435 item->data2 = data2;
436 item->data3 = data3;
437
438 X_LOCK (eq_lock);
439
440 *(eq_head ? &eq_tail->next : &eq_head) = item;
441 eq_tail = item;
442
443 X_UNLOCK (eq_lock);
444
445 eq_signal_func (eq_signal_arg, 0);
446}
447
448static eq_item *
449eq_dec (void)
450{
451 eq_item *res;
452
453 X_LOCK (eq_lock);
454
455 res = eq_head;
456 if (res)
457 eq_head = res->next;
458
459 X_UNLOCK (eq_lock);
460
461 return res;
462}
463
464static void
465eq_poll (void)
466{
467 eq_item *item;
468
469 while ((item = eq_dec ()))
470 {
471 ENTER;
472 SAVETMPS;
473
474 dSP;
475 PUSHMARK (SP);
476 EXTEND (SP, 2);
477
478 if (item->vtbl->free_cb)
479 sv_2mortal (item->cb);
480
481 PUTBACK;
482 item->vtbl->push (item->data1, item->data2, item->data3);
483
484 SV *cb = item->cb;
485 free (item);
486
487 call_sv (cb, G_DISCARD | G_VOID);
488
489 FREETMPS;
490 LEAVE;
491 }
492}
493
494static void
495eq_poll_interrupt (pTHX_ void *c_arg, int value)
496{
497 eq_poll ();
498}
499
500/*****************************************************************************/
501/* context notify */
502
503static void ecb_noinline
504eq_context_push (void *data1, void *data2, void *data3)
505{
506 dSP;
507 PUSHs (sv_2mortal (newSVpv (data1, 0)));
508 PUSHs (sv_2mortal (newSVpvn (data2, (STRLEN)data3)));
509 PUTBACK;
510
511 free (data1);
512 free (data2);
513}
514
515static eq_vtbl eq_context_vtbl = { 0, eq_context_push };
516
517static void CL_CALLBACK
518eq_context_notify (const char *msg, const void *pvt, size_t cb, void *user_data)
519{
520 void *pvt_copy = malloc (cb);
521 memcpy (pvt_copy, pvt, cb);
522 eq_enq (&eq_context_vtbl, user_data, strdup (msg), pvt_copy, (void *)cb);
523}
524
525#define CONTEXT_NOTIFY_CALLBACK \
526 void (CL_CALLBACK *pfn_notify)(const char *, const void *, size_t, void *) = context_default_notify; \
527 void *user_data = 0; \
528 \
529 if (SvOK (notify)) \
530 { \
531 pfn_notify = eq_context_notify; \
532 user_data = s_get_cv (notify); \
533 }
534
535static SV * ecb_noinline
536new_clobj_context (cl_context ctx, void *user_data)
537{
538 SV *sv = new_clobj (stash_context, (IV)ctx);
539
540 if (user_data)
541 sv_magicext (SvRV (sv), user_data, PERL_MAGIC_ext, 0, 0, 0);
542
543 return sv;
544}
545
546#define XPUSH_CLOBJ_CONTEXT XPUSHs (new_clobj_context (ctx, user_data));
547
548/*****************************************************************************/
549/* build/compile/link notify */
550
551static void
552eq_program_push (void *data1, void *data2, void *data3)
553{
554 dSP;
555 PUSH_CLOBJ (stash_program, data1);
556 PUTBACK;
557}
558
559static eq_vtbl eq_program_vtbl = { 1, eq_program_push };
560
561static void CL_CALLBACK
562eq_program_notify (cl_program program, void *user_data)
563{
564 clRetainProgram (program);
565
566 eq_enq (&eq_program_vtbl, user_data, (void *)program, 0, 0);
567}
568
569typedef void (CL_CALLBACK *program_callback)(cl_program program, void *user_data);
570
571static program_callback ecb_noinline
572make_program_callback (SV *notify, void **ruser_data)
573{
574 if (SvOK (notify))
575 {
576 *ruser_data = SvREFCNT_inc (s_get_cv (notify));
577 return eq_program_notify;
578 }
579 else
580 {
581 *ruser_data = 0;
582 return 0;
583 }
584}
585
586struct build_args
587{
588 cl_program program;
589 char *options;
590 void *user_data;
591 cl_uint num_devices;
592};
593
594X_THREAD_PROC (build_program_thread)
595{
596 struct build_args *arg = thr_arg;
597
598 clBuildProgram (arg->program, arg->num_devices, arg->num_devices ? (void *)(arg + 1) : 0, arg->options, 0, 0);
599
600 if (arg->user_data)
601 eq_program_notify (arg->program, arg->user_data);
602 else
603 clReleaseProgram (arg->program);
604
605 free (arg->options);
606 free (arg);
607
608 return 0;
609}
610
611static void
612build_program_async (cl_program program, cl_uint num_devices, const cl_device_id *device_list, const char *options, void *user_data)
613{
614 struct build_args *arg = malloc (sizeof (struct build_args) + sizeof (*device_list) * num_devices);
615
616 arg->program = program;
617 arg->options = strdup (options);
618 arg->user_data = user_data;
619 arg->num_devices = num_devices;
620 memcpy (arg + 1, device_list, sizeof (*device_list) * num_devices);
621
622 xthread_t id;
623 thread_create (&id, build_program_thread, arg);
624}
625
626/*****************************************************************************/
627/* mem object destructor notify */
628
629static void ecb_noinline
630eq_destructor_push (void *data1, void *data2, void *data3)
631{
632}
633
634static eq_vtbl eq_destructor_vtbl = { 0, eq_destructor_push };
635
636static void CL_CALLBACK
637eq_destructor_notify (cl_mem memobj, void *user_data)
638{
639 eq_enq (&eq_destructor_vtbl, (SV *)user_data, (void *)memobj, 0, 0);
640}
641
642/*****************************************************************************/
643/* event objects */
644
645static void
646eq_event_push (void *data1, void *data2, void *data3)
647{
648 dSP;
649 PUSH_CLOBJ (stash_event, data1);
650 PUSHs (sv_2mortal (newSViv ((IV)data2)));
651 PUTBACK;
652}
653
654static eq_vtbl eq_event_vtbl = { 1, eq_event_push };
655
656static void CL_CALLBACK
657eq_event_notify (cl_event event, cl_int event_command_exec_status, void *user_data)
658{
659 clRetainEvent (event);
660 eq_enq (&eq_event_vtbl, user_data, (void *)event, (void *)(IV)event_command_exec_status, 0);
661}
662
663/*****************************************************************************/
664/* utilities for XS code */
262 665
263static size_t 666static size_t
264img_row_pitch (cl_mem img) 667img_row_pitch (cl_mem img)
265{ 668{
266 size_t res; 669 size_t res;
267 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0); 670 clGetImageInfo (img, CL_IMAGE_ROW_PITCH, sizeof (res), &res, 0);
268 return res; 671 return res;
269} 672}
270 673
271static cl_event * 674static cl_event * ecb_noinline
272event_list (SV **items, cl_uint *rcount) 675event_list (CV *cv, SV **items, cl_uint *rcount, cl_event extra)
273{ 676{
274 cl_uint count = *rcount; 677 cl_uint count = *rcount;
275 678
276 if (!count) 679 if (count > 0x7fffffffU) // yeah, it's a hack - the caller might have underflowed
680 *rcount = count = 0;
681
682 if (!count && !extra)
277 return 0; 683 return 0;
278 684
279 cl_event *list = tmpbuf (sizeof (cl_event) * count); 685 cl_event *list = tmpbuf (sizeof (cl_event) * (count + 1));
280 int i = 0; 686 int i = 0;
281 687
282 do 688 while (count--)
283 {
284 --count;
285 if (SvOK (items [count])) 689 if (SvOK (items [count]))
286 list [i++] = SvPTROBJ ("clEnqueue", "wait_events", items [count], "OpenCL::Event"); 690 list [i++] = SvCLOBJ (cv, "wait_events", items [count], "OpenCL::Event");
287 } 691
288 while (count); 692 if (extra)
693 list [i++] = extra;
289 694
290 *rcount = i; 695 *rcount = i;
291 696
292 return i ? list : 0; 697 return i ? list : 0;
293} 698}
294 699
295#define EVENT_LIST(items,count) \ 700#define EVENT_LIST(skip) \
296 cl_uint event_list_count = (count); \ 701 cl_uint event_list_count = items - (skip); \
297 cl_event *event_list_ptr = event_list (&ST (items), &event_list_count) 702 cl_event *event_list_ptr = event_list (cv, &ST (skip), &event_list_count, 0)
298 703
299#define INFO(class) \ 704#define INFO(class) \
300{ \ 705{ \
301 size_t size; \ 706 size_t size; \
302 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \ 707 NEED_SUCCESS (Get ## class ## Info, (self, name, 0, 0, &size)); \
306 SvCUR_set (sv, size); \ 711 SvCUR_set (sv, size); \
307 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \ 712 NEED_SUCCESS (Get ## class ## Info, (self, name, size, SvPVX (sv), 0)); \
308 XPUSHs (sv); \ 713 XPUSHs (sv); \
309} 714}
310 715
716/*****************************************************************************/
717/* mapped_xxx */
718
719static OpenCL__Mapped
720SvMAPPED (SV *sv)
721{
722 // no typechecking atm., keep your fingers crossed
723 return (OpenCL__Mapped)SvMAGIC (SvRV (sv))->mg_ptr;
724}
725
726struct mapped
727{
728 cl_command_queue queue;
729 cl_mem memobj;
730 void *ptr;
731 size_t cb;
732 cl_event event;
733 size_t row_pitch;
734 size_t slice_pitch;
735
736 size_t element_size;
737 size_t width, height, depth;
738};
739
740static SV *
741mapped_new (
742 HV *stash, cl_command_queue queue, cl_mem memobj, cl_map_flags flags,
743 void *ptr, size_t cb, cl_event ev,
744 size_t row_pitch, size_t slice_pitch, size_t element_size,
745 size_t width, size_t height, size_t depth
746)
747{
748 SV *data = newSV (0);
749 SvUPGRADE (data, SVt_PVMG);
750
751 OpenCL__Mapped mapped;
752 New (0, mapped, 1, struct mapped);
753
754 clRetainCommandQueue (queue);
755
756 mapped->queue = queue;
757 mapped->memobj = memobj;
758 mapped->ptr = ptr;
759 mapped->cb = cb;
760 mapped->event = ev;
761 mapped->row_pitch = row_pitch;
762 mapped->slice_pitch = slice_pitch;
763
764 mapped->element_size = element_size;
765 mapped->width = width;
766 mapped->height = height;
767 mapped->depth = depth;
768
769 sv_magicext (data, 0, PERL_MAGIC_ext, 0, (char *)mapped, 0);
770
771 if (SvLEN (data))
772 Safefree (data);
773
774 SvPVX (data) = (char *)ptr;
775 SvCUR_set (data, cb);
776 SvLEN_set (data, 0);
777 SvPOK_only (data);
778
779 SV *obj = sv_2mortal (sv_bless (newRV_noinc (data), stash));
780
781 if (!(flags & CL_MAP_WRITE))
782 SvREADONLY_on (data);
783
784 return obj;
785}
786
787static void
788mapped_detach (SV *sv, OpenCL__Mapped mapped)
789{
790 SV *data = SvRV (sv);
791
792 // the next check checks both before AND after detach, where SvPVX should STILL be 0
793 if (SvPVX (data) != (char *)mapped->ptr)
794 warn ("FATAL: OpenCL memory mapped scalar changed location, detected");
795 else
796 {
797 SvREADONLY_off (data);
798 SvCUR_set (data, 0);
799 SvPVX (data) = 0;
800 SvOK_off (data);
801 }
802
803 mapped->ptr = 0;
804}
805
806static void
807mapped_unmap (CV *cv, SV *self, OpenCL__Mapped mapped, cl_command_queue queue, SV **wait_list, cl_uint event_list_count)
808{
809 cl_event *event_list_ptr = event_list (cv, wait_list, &event_list_count, mapped->event);
810 cl_event ev;
811
812 NEED_SUCCESS (EnqueueUnmapMemObject, (queue, mapped->memobj, mapped->ptr, event_list_count, event_list_ptr, &ev));
813
814 clReleaseEvent (mapped->event);
815 mapped->event = ev;
816
817 mapped_detach (self, mapped);
818}
819
820static size_t
821mapped_element_size (OpenCL__Mapped self)
822{
823 if (!self->element_size)
824 clGetImageInfo (self->memobj, CL_IMAGE_ELEMENT_SIZE, sizeof (self->element_size), &self->element_size, 0);
825
826 return self->element_size;
827}
828
829/*****************************************************************************/
830
311MODULE = OpenCL PACKAGE = OpenCL 831MODULE = OpenCL PACKAGE = OpenCL
312 832
313PROTOTYPES: ENABLE 833PROTOTYPES: ENABLE
314 834
835void
836poll ()
837 CODE:
838 eq_poll ();
839
840void
841_eq_initialise (IV func, IV arg)
842 CODE:
843 eq_signal_func = (void (*)(void *, int))func;
844 eq_signal_arg = (void*)arg;
845
315BOOT: 846BOOT:
316{ 847{
317 HV *stash = gv_stashpv ("OpenCL", 1); 848 HV *stash = gv_stashpv ("OpenCL", 1);
849
318 static const ivstr *civ, const_iv[] = { 850 static const ivstr *civ, const_iv[] = {
319 { sizeof (cl_char ), "SIZEOF_CHAR" }, 851 { sizeof (cl_char ), "SIZEOF_CHAR" },
320 { sizeof (cl_uchar ), "SIZEOF_UCHAR" }, 852 { sizeof (cl_uchar ), "SIZEOF_UCHAR" },
321 { sizeof (cl_short ), "SIZEOF_SHORT" }, 853 { sizeof (cl_short ), "SIZEOF_SHORT" },
322 { sizeof (cl_ushort), "SIZEOF_USHORT" }, 854 { sizeof (cl_ushort), "SIZEOF_USHORT" },
325 { sizeof (cl_long ), "SIZEOF_LONG" }, 857 { sizeof (cl_long ), "SIZEOF_LONG" },
326 { sizeof (cl_ulong ), "SIZEOF_ULONG" }, 858 { sizeof (cl_ulong ), "SIZEOF_ULONG" },
327 { sizeof (cl_half ), "SIZEOF_HALF" }, 859 { sizeof (cl_half ), "SIZEOF_HALF" },
328 { sizeof (cl_float ), "SIZEOF_FLOAT" }, 860 { sizeof (cl_float ), "SIZEOF_FLOAT" },
329 { sizeof (cl_double), "SIZEOF_DOUBLE" }, 861 { sizeof (cl_double), "SIZEOF_DOUBLE" },
862 { PREFER_1_1 , "PREFER_1_1" },
330#include "constiv.h" 863#include "constiv.h"
331 }; 864 };
865
332 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--) 866 for (civ = const_iv + sizeof (const_iv) / sizeof (const_iv [0]); civ > const_iv; civ--)
333 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv)); 867 newCONSTSUB (stash, (char *)civ[-1].name, newSViv (civ[-1].iv));
868
869 static const nvstr *cnv, const_nv[] = {
870#include "constnv.h"
871 };
872
873 for (cnv = const_nv + sizeof (const_nv) / sizeof (const_nv [0]); cnv > const_nv; cnv--)
874 newCONSTSUB (stash, (char *)cnv[-1].name, newSVnv (cnv[-1].nv));
875
876 newCONSTSUB (stash, "NAN", newSVnv (CL_NAN)); // CL_NAN might be a function call
877
878 stash_platform = gv_stashpv ("OpenCL::Platform", GV_ADD);
879 stash_device = gv_stashpv ("OpenCL::Device", GV_ADD);
880 stash_subdevice = gv_stashpv ("OpenCL::SubDevice", GV_ADD);
881 stash_context = gv_stashpv ("OpenCL::Context", GV_ADD);
882 stash_queue = gv_stashpv ("OpenCL::Queue", GV_ADD);
883 stash_program = gv_stashpv ("OpenCL::Program", GV_ADD);
884 stash_kernel = gv_stashpv ("OpenCL::Kernel", GV_ADD);
885 stash_sampler = gv_stashpv ("OpenCL::Sampler", GV_ADD);
886 stash_event = gv_stashpv ("OpenCL::Event", GV_ADD);
887 stash_userevent = gv_stashpv ("OpenCL::UserEvent", GV_ADD);
888 stash_memory = gv_stashpv ("OpenCL::Memory", GV_ADD);
889 stash_buffer = gv_stashpv ("OpenCL::Buffer", GV_ADD);
890 stash_bufferobj = gv_stashpv ("OpenCL::BufferObj", GV_ADD);
891 stash_image = gv_stashpv ("OpenCL::Image", GV_ADD);
892 stash_image1d = gv_stashpv ("OpenCL::Image1D", GV_ADD);
893 stash_image1darray = gv_stashpv ("OpenCL::Image1DArray", GV_ADD);
894 stash_image1dbuffer = gv_stashpv ("OpenCL::Image1DBuffer", GV_ADD);
895 stash_image2d = gv_stashpv ("OpenCL::Image2D", GV_ADD);
896 stash_image2darray = gv_stashpv ("OpenCL::Image2DArray", GV_ADD);
897 stash_image3d = gv_stashpv ("OpenCL::Image3D", GV_ADD);
898 stash_mapped = gv_stashpv ("OpenCL::Mapped", GV_ADD);
899 stash_mappedbuffer = gv_stashpv ("OpenCL::MappedBuffer", GV_ADD);
900 stash_mappedimage = gv_stashpv ("OpenCL::MappedImage", GV_ADD);
901
902 sv_setiv (perl_get_sv ("OpenCL::POLL_FUNC", TRUE), (IV)eq_poll_interrupt);
334} 903}
335 904
336cl_int 905cl_int
337errno () 906errno ()
338 CODE: 907 CODE:
339 RETVAL = res; 908 RETVAL = res;
340 OUTPUT: 909 OUTPUT:
341 RETVAL 910 RETVAL
342 911
343const char * 912const char *
344err2str (cl_int err) 913err2str (cl_int err = res)
345 914
346const char * 915const char *
347enum2str (cl_uint value) 916enum2str (cl_uint value)
348 917
349void 918void
357 list = tmpbuf (sizeof (*list) * count); 926 list = tmpbuf (sizeof (*list) * count);
358 NEED_SUCCESS (GetPlatformIDs, (count, list, 0)); 927 NEED_SUCCESS (GetPlatformIDs, (count, list, 0));
359 928
360 EXTEND (SP, count); 929 EXTEND (SP, count);
361 for (i = 0; i < count; ++i) 930 for (i = 0; i < count; ++i)
362 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", list [i])); 931 PUSH_CLOBJ (stash_platform, list [i]);
363 932
364void 933void
365context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 934context_from_type (cl_context_properties *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
366 PPCODE: 935 PPCODE:
936 CONTEXT_NOTIFY_CALLBACK;
367 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, 0, 0, &res)); 937 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (properties, type, pfn_notify, user_data, &res));
368 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 938 XPUSH_CLOBJ_CONTEXT;
369 939
370#if 0
371
372void 940void
373context (cl_context_properties *properties = 0, FUTURE devices, FUTURE notify = 0) 941context (cl_context_properties *properties, SV *devices, SV *notify = &PL_sv_undef)
942 PPCODE:
943 cl_uint device_count;
944 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
945
946 CONTEXT_NOTIFY_CALLBACK;
947 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, device_count, device_list, pfn_notify, user_data, &res));
948 XPUSH_CLOBJ_CONTEXT;
949
950void
951wait_for_events (...)
374 PPCODE: 952 CODE:
375 /* der Gipfel der Kunst */ 953 EVENT_LIST (0);
954 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
376 955
956PROTOTYPES: DISABLE
957
958MODULE = OpenCL PACKAGE = OpenCL::Platform
959
960void
961info (OpenCL::Platform self, cl_platform_info name)
962 PPCODE:
963 INFO (Platform)
964
965void
966unload_compiler (OpenCL::Platform self)
967 CODE:
968#if CL_VERSION_1_2
969 clUnloadPlatformCompiler (self);
377#endif 970#endif
378
379void
380wait_for_events (...)
381 CODE:
382 EVENT_LIST (0, items);
383 NEED_SUCCESS (WaitForEvents, (event_list_count, event_list_ptr));
384
385PROTOTYPES: DISABLE
386
387MODULE = OpenCL PACKAGE = OpenCL::Platform
388
389void
390info (OpenCL::Platform self, cl_platform_info name)
391 PPCODE:
392 INFO (Platform)
393 971
394#BEGIN:platform 972#BEGIN:platform
395 973
396void 974void
397profile (OpenCL::Platform self) 975profile (OpenCL::Platform self)
423 list = tmpbuf (sizeof (*list) * count); 1001 list = tmpbuf (sizeof (*list) * count);
424 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0)); 1002 NEED_SUCCESS (GetDeviceIDs, (self, type, count, list, 0));
425 1003
426 EXTEND (SP, count); 1004 EXTEND (SP, count);
427 for (i = 0; i < count; ++i) 1005 for (i = 0; i < count; ++i)
428 PUSHs (sv_setref_pv (sv_newmortal (), "OpenCL::Device", list [i])); 1006 PUSH_CLOBJ (stash_device, list [i]);
429 1007
430void 1008void
431context (OpenCL::Platform self, cl_context_properties *properties = 0, SV *devices, FUTURE notify = 0) 1009context (OpenCL::Platform self, SV *properties, SV *devices, SV *notify = &PL_sv_undef)
432 PPCODE: 1010 PPCODE:
1011 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1012 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1013
1014 cl_uint device_count;
1015 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1016
1017 CONTEXT_NOTIFY_CALLBACK;
1018 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (props, device_count, device_list, pfn_notify, user_data, &res));
1019 XPUSH_CLOBJ_CONTEXT;
1020
1021void
1022context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, SV *notify = &PL_sv_undef)
1023 PPCODE:
1024 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
1025 cl_context_properties *props = SvCONTEXTPROPERTIES (cv, "properties", properties, extra, 2);
1026
1027 CONTEXT_NOTIFY_CALLBACK;
1028 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, pfn_notify, user_data, &res));
1029 XPUSH_CLOBJ_CONTEXT;
1030
1031MODULE = OpenCL PACKAGE = OpenCL::Device
1032
1033void
1034info (OpenCL::Device self, cl_device_info name)
1035 PPCODE:
1036 INFO (Device)
1037
1038#if CL_VERSION_1_2
1039
1040void
1041sub_devices (OpenCL::Device self, SV *properties)
1042 PPCODE:
433 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVAV) 1043 if (!SvROK (properties) || SvTYPE (SvRV (properties)) != SVt_PVAV)
434 croak ("OpenCL::Platform::context argument 'device' must be an arrayref with device objects, in call"); 1044 croak ("OpenCL::Device::sub_devices: properties must be specified as reference to an array of property-value pairs");
435 1045
436 AV *av = (AV *)SvRV (devices); 1046 properties = SvRV (properties);
437 cl_uint num_devices = av_len (av) + 1; 1047
438 cl_device_id *device_list = tmpbuf (sizeof (cl_device_id) * num_devices); 1048 cl_uint count = av_len ((AV *)properties) + 1;
1049 cl_device_partition_property *props = tmpbuf (sizeof (*props) * count + 1);
1050
439 int i; 1051 int i;
1052 for (i = 0; i < count; ++i)
1053 props [i] = (cl_device_partition_property)SvIV (*av_fetch ((AV *)properties, i, 0));
440 1054
441 for (i = num_devices; i--; ) 1055 props [count] = 0;
442 device_list [i] = SvPTROBJ ("clCreateContext", "devices", *av_fetch (av, i, 0), "OpenCL::Device");
443 1056
444 NEED_SUCCESS_ARG (cl_context ctx, CreateContext, (properties, num_devices, device_list, 0, 0, &res)); 1057 cl_uint num_devices;
445 XPUSH_NEW_OBJ ("OpenCL::Context", ctx); 1058 NEED_SUCCESS (CreateSubDevices, (self, props, 0, 0, &num_devices));
1059 cl_device_id *list = tmpbuf (sizeof (*list) * num_devices);
1060 NEED_SUCCESS (CreateSubDevices, (self, props, num_devices, list, 0));
446 1061
447void 1062 EXTEND (SP, num_devices);
448context_from_type (OpenCL::Platform self, SV *properties = 0, cl_device_type type = CL_DEVICE_TYPE_DEFAULT, FUTURE notify = 0) 1063 for (i = 0; i < count; ++i)
449 PPCODE: 1064 PUSH_CLOBJ (stash_subdevice, list [i]);
450 cl_context_properties extra[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)self };
451 cl_context_properties *props = SvCONTEXTPROPERTIES ("OpenCL::Platform::context_from_type", "properties", properties, extra, 2);
452 NEED_SUCCESS_ARG (cl_context ctx, CreateContextFromType, (props, type, 0, 0, &res));
453 XPUSH_NEW_OBJ ("OpenCL::Context", ctx);
454 1065
455MODULE = OpenCL PACKAGE = OpenCL::Device 1066#endif
456
457void
458info (OpenCL::Device self, cl_device_info name)
459 PPCODE:
460 INFO (Device)
461 1067
462#BEGIN:device 1068#BEGIN:device
463 1069
464void 1070void
465type (OpenCL::Device self) 1071type (OpenCL::Device self)
1072 ALIAS:
1073 type = CL_DEVICE_TYPE
1074 address_bits = CL_DEVICE_ADDRESS_BITS
1075 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
1076 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
1077 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
1078 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
1079 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
1080 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
1081 execution_capabilities = CL_DEVICE_EXECUTION_CAPABILITIES
1082 properties = CL_DEVICE_QUEUE_PROPERTIES
1083 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
1084 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
466 PPCODE: 1085 PPCODE:
467 cl_device_type value [1]; 1086 cl_ulong value [1];
468 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_TYPE, sizeof (value), value, 0)); 1087 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
469 EXTEND (SP, 1); 1088 EXTEND (SP, 1);
470 const int i = 0; 1089 const int i = 0;
471 PUSHs (sv_2mortal (newSViv (value [i]))); 1090 PUSHs (sv_2mortal (newSVuv (value [i])));
472 1091
473void 1092void
474vendor_id (OpenCL::Device self) 1093vendor_id (OpenCL::Device self)
475 ALIAS: 1094 ALIAS:
476 vendor_id = CL_DEVICE_VENDOR_ID 1095 vendor_id = CL_DEVICE_VENDOR_ID
487 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS 1106 max_write_image_args = CL_DEVICE_MAX_WRITE_IMAGE_ARGS
488 image_support = CL_DEVICE_IMAGE_SUPPORT 1107 image_support = CL_DEVICE_IMAGE_SUPPORT
489 max_samplers = CL_DEVICE_MAX_SAMPLERS 1108 max_samplers = CL_DEVICE_MAX_SAMPLERS
490 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN 1109 mem_base_addr_align = CL_DEVICE_MEM_BASE_ADDR_ALIGN
491 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE 1110 min_data_type_align_size = CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE
1111 global_mem_cache_type = CL_DEVICE_GLOBAL_MEM_CACHE_TYPE
492 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE 1112 global_mem_cacheline_size = CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE
493 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS 1113 max_constant_args = CL_DEVICE_MAX_CONSTANT_ARGS
1114 local_mem_type = CL_DEVICE_LOCAL_MEM_TYPE
494 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF 1115 preferred_vector_width_half = CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF
495 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR 1116 native_vector_width_char = CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR
496 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT 1117 native_vector_width_short = CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT
497 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT 1118 native_vector_width_int = CL_DEVICE_NATIVE_VECTOR_WIDTH_INT
498 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG 1119 native_vector_width_long = CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG
499 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT 1120 native_vector_width_float = CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT
500 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE 1121 native_vector_width_double = CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE
501 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF 1122 native_vector_width_half = CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF
502 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT 1123 reference_count_ext = CL_DEVICE_REFERENCE_COUNT_EXT
503 PPCODE: 1124 PPCODE:
504 cl_uint value [1]; 1125 cl_uint value [1];
505 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1126 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
506 EXTEND (SP, 1); 1127 EXTEND (SP, 1);
507 const int i = 0; 1128 const int i = 0;
536 EXTEND (SP, n); 1157 EXTEND (SP, n);
537 for (i = 0; i < n; ++i) 1158 for (i = 0; i < n; ++i)
538 PUSHs (sv_2mortal (newSVuv (value [i]))); 1159 PUSHs (sv_2mortal (newSVuv (value [i])));
539 1160
540void 1161void
541address_bits (OpenCL::Device self)
542 PPCODE:
543 cl_bitfield value [1];
544 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_ADDRESS_BITS, sizeof (value), value, 0));
545 EXTEND (SP, 1);
546 const int i = 0;
547 PUSHs (sv_2mortal (newSVuv (value [i])));
548
549void
550max_mem_alloc_size (OpenCL::Device self)
551 ALIAS:
552 max_mem_alloc_size = CL_DEVICE_MAX_MEM_ALLOC_SIZE
553 global_mem_cache_size = CL_DEVICE_GLOBAL_MEM_CACHE_SIZE
554 global_mem_size = CL_DEVICE_GLOBAL_MEM_SIZE
555 max_constant_buffer_size = CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE
556 local_mem_size = CL_DEVICE_LOCAL_MEM_SIZE
557 PPCODE:
558 cl_ulong value [1];
559 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
560 EXTEND (SP, 1);
561 const int i = 0;
562 PUSHs (sv_2mortal (newSVuv (value [i])));
563
564void
565single_fp_config (OpenCL::Device self)
566 ALIAS:
567 single_fp_config = CL_DEVICE_SINGLE_FP_CONFIG
568 double_fp_config = CL_DEVICE_DOUBLE_FP_CONFIG
569 half_fp_config = CL_DEVICE_HALF_FP_CONFIG
570 PPCODE:
571 cl_device_fp_config value [1];
572 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
573 EXTEND (SP, 1);
574 const int i = 0;
575 PUSHs (sv_2mortal (newSVuv (value [i])));
576
577void
578global_mem_cache_type (OpenCL::Device self)
579 PPCODE:
580 cl_device_mem_cache_type value [1];
581 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE, sizeof (value), value, 0));
582 EXTEND (SP, 1);
583 const int i = 0;
584 PUSHs (sv_2mortal (newSVuv (value [i])));
585
586void
587local_mem_type (OpenCL::Device self)
588 PPCODE:
589 cl_device_local_mem_type value [1];
590 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_LOCAL_MEM_TYPE, sizeof (value), value, 0));
591 EXTEND (SP, 1);
592 const int i = 0;
593 PUSHs (sv_2mortal (newSVuv (value [i])));
594
595void
596error_correction_support (OpenCL::Device self) 1162error_correction_support (OpenCL::Device self)
597 ALIAS: 1163 ALIAS:
598 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT 1164 error_correction_support = CL_DEVICE_ERROR_CORRECTION_SUPPORT
599 endian_little = CL_DEVICE_ENDIAN_LITTLE 1165 endian_little = CL_DEVICE_ENDIAN_LITTLE
600 available = CL_DEVICE_AVAILABLE 1166 available = CL_DEVICE_AVAILABLE
601 compiler_available = CL_DEVICE_COMPILER_AVAILABLE 1167 compiler_available = CL_DEVICE_COMPILER_AVAILABLE
602 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY 1168 host_unified_memory = CL_DEVICE_HOST_UNIFIED_MEMORY
603 PPCODE: 1169 PPCODE:
604 cl_bool value [1]; 1170 cl_uint value [1];
605 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0)); 1171 NEED_SUCCESS (GetDeviceInfo, (self, ix, sizeof (value), value, 0));
606 EXTEND (SP, 1); 1172 EXTEND (SP, 1);
607 const int i = 0; 1173 const int i = 0;
608 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 1174 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
609
610void
611execution_capabilities (OpenCL::Device self)
612 PPCODE:
613 cl_device_exec_capabilities value [1];
614 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_EXECUTION_CAPABILITIES, sizeof (value), value, 0));
615 EXTEND (SP, 1);
616 const int i = 0;
617 PUSHs (sv_2mortal (newSVuv (value [i])));
618
619void
620properties (OpenCL::Device self)
621 PPCODE:
622 cl_command_queue_properties value [1];
623 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_QUEUE_PROPERTIES, sizeof (value), value, 0));
624 EXTEND (SP, 1);
625 const int i = 0;
626 PUSHs (sv_2mortal (newSViv (value [i])));
627 1175
628void 1176void
629platform (OpenCL::Device self) 1177platform (OpenCL::Device self)
630 PPCODE: 1178 PPCODE:
631 cl_platform_id value [1]; 1179 cl_platform_id value [1];
632 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0)); 1180 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PLATFORM, sizeof (value), value, 0));
633 EXTEND (SP, 1); 1181 EXTEND (SP, 1);
634 const int i = 0; 1182 const int i = 0;
635 { 1183 PUSH_CLOBJ (stash_platform, value [i]);
636 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Platform", value [i]));
637 }
638 1184
639void 1185void
640name (OpenCL::Device self) 1186name (OpenCL::Device self)
641 ALIAS: 1187 ALIAS:
642 name = CL_DEVICE_NAME 1188 name = CL_DEVICE_NAME
659 PPCODE: 1205 PPCODE:
660 cl_device_id value [1]; 1206 cl_device_id value [1];
661 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0)); 1207 NEED_SUCCESS (GetDeviceInfo, (self, CL_DEVICE_PARENT_DEVICE_EXT, sizeof (value), value, 0));
662 EXTEND (SP, 1); 1208 EXTEND (SP, 1);
663 const int i = 0; 1209 const int i = 0;
664 { 1210 PUSH_CLOBJ (stash_device, value [i]);
665 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
666 }
667 1211
668void 1212void
669partition_types_ext (OpenCL::Device self) 1213partition_types_ext (OpenCL::Device self)
670 ALIAS: 1214 ALIAS:
671 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT 1215 partition_types_ext = CL_DEVICE_PARTITION_TYPES_EXT
681 for (i = 0; i < n; ++i) 1225 for (i = 0; i < n; ++i)
682 PUSHs (sv_2mortal (newSVuv (value [i]))); 1226 PUSHs (sv_2mortal (newSVuv (value [i])));
683 1227
684#END:device 1228#END:device
685 1229
1230MODULE = OpenCL PACKAGE = OpenCL::SubDevice
1231
1232#if CL_VERSION_1_2
1233
1234void
1235DESTROY (OpenCL::SubDevice self)
1236 CODE:
1237 clReleaseDevice (self);
1238
1239#endif
1240
686MODULE = OpenCL PACKAGE = OpenCL::Context 1241MODULE = OpenCL PACKAGE = OpenCL::Context
687 1242
688void 1243void
689DESTROY (OpenCL::Context context) 1244DESTROY (OpenCL::Context self)
690 CODE: 1245 CODE:
691 clReleaseContext (context); 1246 clReleaseContext (self);
692 1247
693void 1248void
694info (OpenCL::Context self, cl_context_info name) 1249info (OpenCL::Context self, cl_context_info name)
695 PPCODE: 1250 PPCODE:
696 INFO (Context) 1251 INFO (Context)
697 1252
698void 1253void
699queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0) 1254queue (OpenCL::Context self, OpenCL::Device device, cl_command_queue_properties properties = 0)
700 PPCODE: 1255 PPCODE:
701 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res)); 1256 NEED_SUCCESS_ARG (cl_command_queue queue, CreateCommandQueue, (self, device, properties, &res));
702 XPUSH_NEW_OBJ ("OpenCL::Queue", queue); 1257 XPUSH_CLOBJ (stash_queue, queue);
703 1258
704void 1259void
705user_event (OpenCL::Context self) 1260user_event (OpenCL::Context self)
706 PPCODE: 1261 PPCODE:
707 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res)); 1262 NEED_SUCCESS_ARG (cl_event ev, CreateUserEvent, (self, &res));
708 XPUSH_NEW_OBJ ("OpenCL::UserEvent", ev); 1263 XPUSH_CLOBJ (stash_userevent, ev);
709 1264
710void 1265void
711buffer (OpenCL::Context self, cl_mem_flags flags, size_t len) 1266buffer (OpenCL::Context self, cl_mem_flags flags, size_t len)
712 PPCODE: 1267 PPCODE:
713 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)) 1268 if (flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))
714 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?"); 1269 croak ("OpenCL::Context::buffer: cannot use/copy host ptr when no data is given, use $context->buffer_sv instead?");
715 1270
716 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res)); 1271 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, 0, &res));
717 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1272 XPUSH_CLOBJ (stash_bufferobj, mem);
718 1273
719void 1274void
720buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data) 1275buffer_sv (OpenCL::Context self, cl_mem_flags flags, SV *data)
721 PPCODE: 1276 PPCODE:
722 STRLEN len; 1277 STRLEN len;
723 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1278 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
724 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR))) 1279 if (!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
725 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?"); 1280 croak ("OpenCL::Context::buffer_sv: you have to specify use or copy host ptr when buffer data is given, use $context->buffer instead?");
726 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res)); 1281 NEED_SUCCESS_ARG (cl_mem mem, CreateBuffer, (self, flags, len, ptr, &res));
727 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1282 XPUSH_CLOBJ (stash_bufferobj, mem);
728 1283
729#if !defined CL_VERSION_1_2 || defined CL_USE_DEPRECATED_OPENCL_1_1_APIS 1284#if CL_VERSION_1_2
730 1285
731void 1286void
732image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef) 1287image (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, cl_mem_object_type type, size_t width, size_t height, size_t depth = 0, size_t array_size = 0, size_t row_pitch = 0, size_t slice_pitch = 0, cl_uint num_mip_level = 0, cl_uint num_samples = 0, SV *data = &PL_sv_undef)
733 PPCODE: 1288 PPCODE:
734 STRLEN len; 1289 STRLEN len;
735 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1290 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
736 const cl_image_format format = { channel_order, channel_type }; 1291 const cl_image_format format = { channel_order, channel_type };
1292 const cl_image_desc desc = {
1293 type,
1294 width, height, depth,
1295 array_size, row_pitch, slice_pitch,
1296 num_mip_level, num_samples,
1297 type == CL_MEM_OBJECT_IMAGE1D_BUFFER ? (cl_mem)SvCLOBJ (cv, "data", data, "OpenCL::Buffer") : 0
1298 };
737 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res)); 1299 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
738 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1300 HV *stash = stash_image;
1301 switch (type)
1302 {
1303 case CL_MEM_OBJECT_IMAGE1D_BUFFER: stash = stash_image1dbuffer; break;
1304 case CL_MEM_OBJECT_IMAGE1D: stash = stash_image1d; break;
1305 case CL_MEM_OBJECT_IMAGE1D_ARRAY: stash = stash_image2darray; break;
1306 case CL_MEM_OBJECT_IMAGE2D: stash = stash_image2d; break;
1307 case CL_MEM_OBJECT_IMAGE2D_ARRAY: stash = stash_image2darray; break;
1308 case CL_MEM_OBJECT_IMAGE3D: stash = stash_image3d; break;
1309 }
1310 XPUSH_CLOBJ (stash, mem);
739 1311
1312#endif
1313
740void 1314void
741image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef) 1315image2d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t row_pitch = 0, SV *data = &PL_sv_undef)
742 PPCODE: 1316 PPCODE:
743 STRLEN len; 1317 STRLEN len;
744 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0; 1318 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
745 const cl_image_format format = { channel_order, channel_type }; 1319 const cl_image_format format = { channel_order, channel_type };
1320#if PREFER_1_1
1321 NEED_SUCCESS_ARG (cl_mem mem, CreateImage2D, (self, flags, &format, width, height, row_pitch, ptr, &res));
1322#else
1323 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE2D, width, height, 0, 0, row_pitch, 0, 0, 0, 0 };
1324 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
1325#endif
1326 XPUSH_CLOBJ (stash_image2d, mem);
1327
1328void
1329image3d (OpenCL::Context self, cl_mem_flags flags, cl_channel_order channel_order, cl_channel_type channel_type, size_t width, size_t height, size_t depth, size_t row_pitch = 0, size_t slice_pitch = 0, SV *data = &PL_sv_undef)
1330 PPCODE:
1331 STRLEN len;
1332 char *ptr = SvOK (data) ? SvPVbyte (data, len) : 0;
1333 const cl_image_format format = { channel_order, channel_type };
1334#if PREFER_1_1
746 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res)); 1335 NEED_SUCCESS_ARG (cl_mem mem, CreateImage3D, (self, flags, &format, width, height, depth, row_pitch, slice_pitch, ptr, &res));
747 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1336#else
748 1337 const cl_image_desc desc = { CL_MEM_OBJECT_IMAGE3D, width, height, depth, 0, row_pitch, slice_pitch, 0, 0, 0 };
1338 NEED_SUCCESS_ARG (cl_mem mem, CreateImage, (self, flags, &format, &desc, ptr, &res));
749#endif 1339#endif
1340 XPUSH_CLOBJ (stash_image3d, mem);
750 1341
751#if cl_apple_gl_sharing || cl_khr_gl_sharing 1342#if cl_apple_gl_sharing || cl_khr_gl_sharing
752 1343
753void 1344void
754gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj) 1345gl_buffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint bufobj)
755 PPCODE: 1346 PPCODE:
756 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res)); 1347 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLBuffer, (self, flags, bufobj, &res));
757 XPUSH_NEW_OBJ ("OpenCL::BufferObj", mem); 1348 XPUSH_CLOBJ (stash_bufferobj, mem);
758 1349
759#if !defined CL_VERSION_1_2 || defined CL_USE_DEPRECATED_OPENCL_1_1_APIS 1350void
1351gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
1352 PPCODE:
1353 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
1354 XPUSH_CLOBJ (stash_image2d, mem);
1355
1356#if CL_VERSION_1_2
1357
1358void
1359gl_texture (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
1360 ALIAS:
1361 PPCODE:
1362 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture, (self, flags, target, miplevel, texture, &res));
1363 cl_gl_object_type type;
1364 NEED_SUCCESS (GetGLObjectInfo, (mem, &type, 0)); // TODO: use target instead?
1365 HV *stash = stash_memory;
1366 switch (type)
1367 {
1368 case CL_GL_OBJECT_TEXTURE_BUFFER: stash = stash_image1dbuffer; break;
1369 case CL_GL_OBJECT_TEXTURE1D: stash = stash_image1d; break;
1370 case CL_GL_OBJECT_TEXTURE1D_ARRAY: stash = stash_image2darray; break;
1371 case CL_GL_OBJECT_TEXTURE2D: stash = stash_image2d; break;
1372 case CL_GL_OBJECT_TEXTURE2D_ARRAY: stash = stash_image2darray; break;
1373 case CL_GL_OBJECT_TEXTURE3D: stash = stash_image3d; break;
1374 }
1375 XPUSH_CLOBJ (stash, mem);
1376
1377#endif
760 1378
761void 1379void
762gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1380gl_texture2d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
763 PPCODE: 1381 PPCODE:
1382#if PREFER_1_1
764 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res)); 1383 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture2D, (self, flags, target, miplevel, texture, &res));
765 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem); 1384#else
1385 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
1386#endif
1387 XPUSH_CLOBJ (stash_image2d, mem);
766 1388
767void 1389void
768gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture) 1390gl_texture3d (OpenCL::Context self, cl_mem_flags flags, cl_GLenum target, cl_GLint miplevel, cl_GLuint texture)
769 PPCODE: 1391 PPCODE:
1392#if PREFER_1_1
770 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res)); 1393 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture3D, (self, flags, target, miplevel, texture, &res));
771 XPUSH_NEW_OBJ ("OpenCL::Image3D", mem); 1394#else
772 1395 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLTexture , (self, flags, target, miplevel, texture, &res));
773#endif 1396#endif
774 1397 XPUSH_CLOBJ (stash_image3d, mem);
775void
776gl_renderbuffer (OpenCL::Context self, cl_mem_flags flags, cl_GLuint renderbuffer)
777 PPCODE:
778 NEED_SUCCESS_ARG (cl_mem mem, CreateFromGLRenderbuffer, (self, flags, renderbuffer, &res));
779 XPUSH_NEW_OBJ ("OpenCL::Image2D", mem);
780 1398
781#endif 1399#endif
782 1400
783void 1401void
784supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type) 1402supported_image_formats (OpenCL::Context self, cl_mem_flags flags, cl_mem_object_type image_type)
804 1422
805void 1423void
806sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode) 1424sampler (OpenCL::Context self, cl_bool normalized_coords, cl_addressing_mode addressing_mode, cl_filter_mode filter_mode)
807 PPCODE: 1425 PPCODE:
808 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res)); 1426 NEED_SUCCESS_ARG (cl_sampler sampler, CreateSampler, (self, normalized_coords, addressing_mode, filter_mode, &res));
809 XPUSH_NEW_OBJ ("OpenCL::Sampler", sampler); 1427 XPUSH_CLOBJ (stash_sampler, sampler);
810 1428
811void 1429void
812program_with_source (OpenCL::Context self, SV *program) 1430program_with_source (OpenCL::Context self, SV *program)
813 PPCODE: 1431 PPCODE:
814 STRLEN len; 1432 STRLEN len;
815 size_t len2; 1433 size_t len2;
816 const char *ptr = SvPVbyte (program, len); 1434 const char *ptr = SvPVbyte (program, len);
817 1435
818 len2 = len; 1436 len2 = len;
819 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res)); 1437 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithSource, (self, 1, &ptr, &len2, &res));
820 XPUSH_NEW_OBJ ("OpenCL::Program", prog); 1438 XPUSH_CLOBJ (stash_program, prog);
1439
1440void
1441program_with_binary (OpenCL::Context self, SV *devices, SV *binaries)
1442 PPCODE:
1443 cl_uint device_count;
1444 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1445
1446 if (!SvROK (binaries) || SvTYPE (SvRV (binaries)) != SVt_PVAV)
1447 croak ("OpenCL::Context::program_with_binary: binaries must be specified as reference to an array of strings");
1448
1449 binaries = SvRV (binaries);
1450
1451 if (device_count != av_len ((AV *)binaries) + 1)
1452 croak ("OpenCL::Context::program_with_binary: differing numbers of devices and binaries are not allowed");
1453
1454 size_t *length_list = tmpbuf (sizeof (*length_list) * device_count);
1455 const unsigned char **binary_list = tmpbuf (sizeof (*binary_list) * device_count);
1456 cl_int *status_list = tmpbuf (sizeof (*status_list) * device_count);
1457
1458 int i;
1459 for (i = 0; i < device_count; ++i)
1460 {
1461 STRLEN len;
1462 binary_list [i] = (const unsigned char *)SvPVbyte (*av_fetch ((AV *)binaries, i, 0), len);
1463 length_list [i] = len;
1464 }
1465
1466 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBinary, (self, device_count, device_list,
1467 length_list, binary_list,
1468 GIMME_V == G_ARRAY ? status_list : 0, &res));
1469
1470 EXTEND (SP, 2);
1471 PUSH_CLOBJ (stash_program, prog);
1472
1473 if (GIMME_V == G_ARRAY)
1474 {
1475 AV *av = newAV ();
1476 PUSHs (sv_2mortal (newRV_noinc ((SV *)av)));
1477
1478 for (i = device_count; i--; )
1479 av_store (av, i, newSViv (status_list [i]));
1480 }
1481
1482#if CL_VERSION_1_2
1483
1484void
1485program_with_built_in_kernels (OpenCL::Context self, SV *devices, SV *kernel_names)
1486 PPCODE:
1487 cl_uint device_count;
1488 cl_device_id *device_list = object_list (cv, 0, "devices", devices, "OpenCL::Device", &device_count);
1489
1490 NEED_SUCCESS_ARG (cl_program prog, CreateProgramWithBuiltInKernels, (self, device_count, device_list, SvPVbyte_nolen (kernel_names), &res));
1491
1492 XPUSH_CLOBJ (stash_program, prog);
1493
1494void
1495link_program (OpenCL::Context self, SV *devices, SV *options, SV *programs, SV *notify = &PL_sv_undef)
1496 CODE:
1497 cl_uint device_count = 0;
1498 cl_device_id *device_list = 0;
1499
1500 if (SvOK (devices))
1501 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
1502
1503 cl_uint program_count;
1504 cl_program *program_list = object_list (cv, 0, "programs", programs, "OpenCL::Program", &program_count);
1505
1506 void *user_data;
1507 program_callback pfn_notify = make_program_callback (notify, &user_data);
1508
1509 NEED_SUCCESS_ARG (cl_program prog, LinkProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
1510 program_count, program_list, pfn_notify, user_data, &res));
1511
1512 XPUSH_CLOBJ (stash_program, prog);
1513
1514#endif
821 1515
822#BEGIN:context 1516#BEGIN:context
823 1517
824void 1518void
825reference_count (OpenCL::Context self) 1519reference_count (OpenCL::Context self)
841 cl_device_id *value = tmpbuf (size); 1535 cl_device_id *value = tmpbuf (size);
842 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0)); 1536 NEED_SUCCESS (GetContextInfo, (self, CL_CONTEXT_DEVICES, size, value, 0));
843 int i, n = size / sizeof (*value); 1537 int i, n = size / sizeof (*value);
844 EXTEND (SP, n); 1538 EXTEND (SP, n);
845 for (i = 0; i < n; ++i) 1539 for (i = 0; i < n; ++i)
846 { 1540 PUSH_CLOBJ (stash_device, value [i]);
847 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
848 }
849 1541
850void 1542void
851properties (OpenCL::Context self) 1543properties (OpenCL::Context self)
852 PPCODE: 1544 PPCODE:
853 size_t size; 1545 size_t size;
867DESTROY (OpenCL::Queue self) 1559DESTROY (OpenCL::Queue self)
868 CODE: 1560 CODE:
869 clReleaseCommandQueue (self); 1561 clReleaseCommandQueue (self);
870 1562
871void 1563void
872enqueue_read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...) 1564read_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, size_t len, SV *data, ...)
1565 ALIAS:
1566 enqueue_read_buffer = 0
873 PPCODE: 1567 PPCODE:
874 cl_event ev = 0;
875 EVENT_LIST (6, items - 6); 1568 EVENT_LIST (6);
876 1569
877 SvUPGRADE (data, SVt_PV); 1570 SvUPGRADE (data, SVt_PV);
878 SvGROW (data, len); 1571 SvGROW (data, len);
879 SvPOK_only (data); 1572 SvPOK_only (data);
880 SvCUR_set (data, len); 1573 SvCUR_set (data, len);
1574
1575 cl_event ev = 0;
881 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1576 NEED_SUCCESS (EnqueueReadBuffer, (self, mem, blocking, offset, len, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
882 1577
883 if (ev) 1578 if (ev)
884 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1579 XPUSH_CLOBJ (stash_event, ev);
885 1580
886void 1581void
887enqueue_write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...) 1582write_buffer (OpenCL::Queue self, OpenCL::Buffer mem, cl_bool blocking, size_t offset, SV *data, ...)
1583 ALIAS:
1584 enqueue_write_buffer = 0
888 PPCODE: 1585 PPCODE:
889 cl_event ev = 0; 1586 EVENT_LIST (5);
1587
890 STRLEN len; 1588 STRLEN len;
891 char *ptr = SvPVbyte (data, len); 1589 char *ptr = SvPVbyte (data, len);
892 EVENT_LIST (5, items - 5);
893 1590
1591 cl_event ev = 0;
894 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1592 NEED_SUCCESS (EnqueueWriteBuffer, (self, mem, blocking, offset, len, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
895 1593
896 if (ev) 1594 if (ev)
897 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1595 XPUSH_CLOBJ (stash_event, ev);
898 1596
1597#if CL_VERSION_1_2
1598
899void 1599void
900enqueue_copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...) 1600fill_buffer (OpenCL::Queue self, OpenCL::Buffer mem, SV *data, size_t offset, size_t size, ...)
1601 ALIAS:
1602 enqueue_fill_buffer = 0
901 PPCODE: 1603 PPCODE:
1604 EVENT_LIST (5);
1605
1606 STRLEN len;
1607 char *ptr = SvPVbyte (data, len);
1608
902 cl_event ev = 0; 1609 cl_event ev = 0;
1610 NEED_SUCCESS (EnqueueFillBuffer, (self, mem, ptr, len, offset, size, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1611
1612 if (ev)
1613 XPUSH_CLOBJ (stash_event, ev);
1614
1615void
1616fill_image (OpenCL::Queue self, OpenCL::Image img, NV r, NV g, NV b, NV a, size_t x, size_t y, size_t z, size_t width, size_t height, size_t depth, ...)
1617 ALIAS:
1618 enqueue_fill_image = 0
1619 PPCODE:
1620 EVENT_LIST (12);
1621
1622 const size_t origin [3] = { x, y, z };
1623 const size_t region [3] = { width, height, depth };
1624
1625 const cl_float c_f [4] = { r, g, b, a };
1626 const cl_uint c_u [4] = { r, g, b, a };
1627 const cl_int c_s [4] = { r, g, b, a };
1628 const void *c_fus [3] = { &c_f, &c_u, &c_s };
1629 static const char fus [] = { 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 1, 1, 1, 0, 0 };
1630 cl_image_format format;
1631 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
1632 assert (sizeof (fus) == CL_FLOAT + 1 - CL_SNORM_INT8);
1633 if (format.image_channel_data_type < CL_SNORM_INT8 || CL_FLOAT < format.image_channel_data_type)
1634 croak ("enqueue_fill_image: image has unsupported channel type, only opencl 1.2 channel types supported.");
1635
1636 cl_event ev = 0;
1637 NEED_SUCCESS (EnqueueFillImage, (self, img, c_fus [fus [format.image_channel_data_type - CL_SNORM_INT8]],
1638 origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1639
1640 if (ev)
1641 XPUSH_CLOBJ (stash_event, ev);
1642
1643#endif
1644
1645void
1646copy_buffer (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_offset, size_t dst_offset, size_t len, ...)
1647 ALIAS:
1648 enqueue_copy_buffer = 0
1649 PPCODE:
903 EVENT_LIST (6, items - 6); 1650 EVENT_LIST (6);
904 1651
1652 cl_event ev = 0;
905 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1653 NEED_SUCCESS (EnqueueCopyBuffer, (self, src, dst, src_offset, dst_offset, len, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
906 1654
907 if (ev) 1655 if (ev)
908 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1656 XPUSH_CLOBJ (stash_event, ev);
909 1657
910void 1658void
911enqueue_read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1659read_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1660 ALIAS:
1661 enqueue_read_buffer_rect = 0
912 PPCODE: 1662 PPCODE:
913 cl_event ev = 0; 1663 EVENT_LIST (17);
1664
914 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1665 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
915 const size_t host_origin[3] = { host_x, host_y, host_z }; 1666 const size_t host_origin[3] = { host_x, host_y, host_z };
916 const size_t region[3] = { width, height, depth }; 1667 const size_t region[3] = { width, height, depth };
917 EVENT_LIST (17, items - 17);
918 1668
919 if (!buf_row_pitch) 1669 if (!buf_row_pitch)
920 buf_row_pitch = region [0]; 1670 buf_row_pitch = region [0];
921 1671
922 if (!buf_slice_pitch) 1672 if (!buf_slice_pitch)
932 1682
933 SvUPGRADE (data, SVt_PV); 1683 SvUPGRADE (data, SVt_PV);
934 SvGROW (data, len); 1684 SvGROW (data, len);
935 SvPOK_only (data); 1685 SvPOK_only (data);
936 SvCUR_set (data, len); 1686 SvCUR_set (data, len);
1687
1688 cl_event ev = 0;
937 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1689 NEED_SUCCESS (EnqueueReadBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
938 1690
939 if (ev) 1691 if (ev)
940 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1692 XPUSH_CLOBJ (stash_event, ev);
941 1693
942void 1694void
943enqueue_write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...) 1695write_buffer_rect (OpenCL::Queue self, OpenCL::Memory buf, cl_bool blocking, size_t buf_x, size_t buf_y, size_t buf_z, size_t host_x, size_t host_y, size_t host_z, size_t width, size_t height, size_t depth, size_t buf_row_pitch, size_t buf_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, SV *data, ...)
1696 ALIAS:
1697 enqueue_write_buffer_rect = 0
944 PPCODE: 1698 PPCODE:
945 cl_event ev = 0; 1699 EVENT_LIST (17);
1700
946 const size_t buf_origin [3] = { buf_x , buf_y , buf_z }; 1701 const size_t buf_origin [3] = { buf_x , buf_y , buf_z };
947 const size_t host_origin[3] = { host_x, host_y, host_z }; 1702 const size_t host_origin[3] = { host_x, host_y, host_z };
948 const size_t region[3] = { width, height, depth }; 1703 const size_t region[3] = { width, height, depth };
949 STRLEN len; 1704 STRLEN len;
950 char *ptr = SvPVbyte (data, len); 1705 char *ptr = SvPVbyte (data, len);
951 EVENT_LIST (17, items - 17);
952 1706
953 if (!buf_row_pitch) 1707 if (!buf_row_pitch)
954 buf_row_pitch = region [0]; 1708 buf_row_pitch = region [0];
955 1709
956 if (!buf_slice_pitch) 1710 if (!buf_slice_pitch)
965 size_t min_len = host_row_pitch * host_slice_pitch * region [2]; 1719 size_t min_len = host_row_pitch * host_slice_pitch * region [2];
966 1720
967 if (len < min_len) 1721 if (len < min_len)
968 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1722 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
969 1723
1724 cl_event ev = 0;
970 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1725 NEED_SUCCESS (EnqueueWriteBufferRect, (self, buf, blocking, buf_origin, host_origin, region, buf_row_pitch, buf_slice_pitch, host_row_pitch, host_slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
971 1726
972 if (ev) 1727 if (ev)
973 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1728 XPUSH_CLOBJ (stash_event, ev);
974 1729
975void 1730void
976enqueue_copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...) 1731copy_buffer_rect (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, ...)
1732 ALIAS:
1733 enqueue_copy_buffer_rect = 0
977 PPCODE: 1734 PPCODE:
978 cl_event ev = 0; 1735 EVENT_LIST (16);
1736
979 const size_t src_origin[3] = { src_x, src_y, src_z }; 1737 const size_t src_origin[3] = { src_x, src_y, src_z };
980 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1738 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
981 const size_t region[3] = { width, height, depth }; 1739 const size_t region[3] = { width, height, depth };
982 EVENT_LIST (16, items - 16);
983 1740
1741 cl_event ev = 0;
984 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1742 NEED_SUCCESS (EnqueueCopyBufferRect, (self, src, dst, src_origin, dst_origin, region, src_row_pitch, src_slice_pitch, dst_row_pitch, dst_slice_pitch, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
985 1743
986 if (ev) 1744 if (ev)
987 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1745 XPUSH_CLOBJ (stash_event, ev);
988 1746
989void 1747void
990enqueue_read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1748read_image (OpenCL::Queue self, OpenCL::Image src, cl_bool blocking, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1749 ALIAS:
1750 enqueue_read_image = 0
991 PPCODE: 1751 PPCODE:
992 cl_event ev = 0; 1752 EVENT_LIST (12);
1753
993 const size_t src_origin[3] = { src_x, src_y, src_z }; 1754 const size_t src_origin[3] = { src_x, src_y, src_z };
994 const size_t region[3] = { width, height, depth }; 1755 const size_t region[3] = { width, height, depth };
995 EVENT_LIST (12, items - 12);
996 1756
997 if (!row_pitch) 1757 if (!row_pitch)
998 row_pitch = img_row_pitch (src); 1758 row_pitch = img_row_pitch (src);
999 1759
1000 if (depth > 1 && !slice_pitch) 1760 if (depth > 1 && !slice_pitch)
1004 1764
1005 SvUPGRADE (data, SVt_PV); 1765 SvUPGRADE (data, SVt_PV);
1006 SvGROW (data, len); 1766 SvGROW (data, len);
1007 SvPOK_only (data); 1767 SvPOK_only (data);
1008 SvCUR_set (data, len); 1768 SvCUR_set (data, len);
1769
1770 cl_event ev = 0;
1009 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1771 NEED_SUCCESS (EnqueueReadImage, (self, src, blocking, src_origin, region, row_pitch, slice_pitch, SvPVX (data), event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1010 1772
1011 if (ev) 1773 if (ev)
1012 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1774 XPUSH_CLOBJ (stash_event, ev);
1013 1775
1014void 1776void
1015enqueue_write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...) 1777write_image (OpenCL::Queue self, OpenCL::Image dst, cl_bool blocking, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, size_t row_pitch, size_t slice_pitch, SV *data, ...)
1778 ALIAS:
1779 enqueue_write_image = 0
1016 PPCODE: 1780 PPCODE:
1017 cl_event ev = 0; 1781 EVENT_LIST (12);
1782
1018 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1783 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1019 const size_t region[3] = { width, height, depth }; 1784 const size_t region[3] = { width, height, depth };
1020 STRLEN len; 1785 STRLEN len;
1021 char *ptr = SvPVbyte (data, len); 1786 char *ptr = SvPVbyte (data, len);
1022 EVENT_LIST (12, items - 12);
1023 1787
1024 if (!row_pitch) 1788 if (!row_pitch)
1025 row_pitch = img_row_pitch (dst); 1789 row_pitch = img_row_pitch (dst);
1026 1790
1027 if (depth > 1 && !slice_pitch) 1791 if (depth > 1 && !slice_pitch)
1030 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height; 1794 size_t min_len = slice_pitch ? slice_pitch * depth : row_pitch * height;
1031 1795
1032 if (len < min_len) 1796 if (len < min_len)
1033 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred"); 1797 croak ("clEnqueueWriteImage: data string is shorter than what would be transferred");
1034 1798
1799 cl_event ev = 0;
1035 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1800 NEED_SUCCESS (EnqueueWriteImage, (self, dst, blocking, dst_origin, region, row_pitch, slice_pitch, ptr, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1036 1801
1037 if (ev) 1802 if (ev)
1038 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1803 XPUSH_CLOBJ (stash_event, ev);
1039 1804
1040void 1805void
1041enqueue_copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...) 1806copy_image (OpenCL::Queue self, OpenCL::Image src, OpenCL::Image dst, size_t src_x, size_t src_y, size_t src_z, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1807 ALIAS:
1808 enqueue_copy_image = 0
1042 PPCODE: 1809 PPCODE:
1043 cl_event ev = 0; 1810 EVENT_LIST (12);
1811
1044 const size_t src_origin[3] = { src_x, src_y, src_z }; 1812 const size_t src_origin[3] = { src_x, src_y, src_z };
1045 const size_t dst_origin[3] = { dst_x, dst_y, dst_z }; 1813 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1046 const size_t region[3] = { width, height, depth }; 1814 const size_t region[3] = { width, height, depth };
1047 EVENT_LIST (12, items - 12);
1048 1815
1816 cl_event ev = 0;
1049 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1817 NEED_SUCCESS (EnqueueCopyImage, (self, src, dst, src_origin, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1050 1818
1051 if (ev) 1819 if (ev)
1052 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1820 XPUSH_CLOBJ (stash_event, ev);
1053 1821
1054void 1822void
1055enqueue_copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...) 1823copy_image_to_buffer (OpenCL::Queue self, OpenCL::Image src, OpenCL::Buffer dst, size_t src_x, size_t src_y, size_t src_z, size_t width, size_t height, size_t depth, size_t dst_offset, ...)
1824 ALIAS:
1825 enqueue_copy_image_to_buffer = 0
1056 PPCODE: 1826 PPCODE:
1827 EVENT_LIST (10);
1828
1829 const size_t src_origin[3] = { src_x, src_y, src_z };
1830 const size_t region [3] = { width, height, depth };
1831
1057 cl_event ev = 0; 1832 cl_event ev = 0;
1833 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1834
1835 if (ev)
1836 XPUSH_CLOBJ (stash_event, ev);
1837
1838void
1839copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1840 ALIAS:
1841 enqueue_copy_buffer_to_image = 0
1842 PPCODE:
1843 EVENT_LIST (10);
1844
1058 const size_t src_origin[3] = { src_x, src_y, src_z }; 1845 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1846 const size_t region [3] = { width, height, depth };
1847
1848 cl_event ev = 0;
1849 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1850
1851 if (ev)
1852 XPUSH_CLOBJ (stash_event, ev);
1853
1854void
1855map_buffer (OpenCL::Queue self, OpenCL::Buffer buf, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t offset = 0, SV *cb_ = &PL_sv_undef, ...)
1856 ALIAS:
1857 enqueue_map_buffer = 0
1858 PPCODE:
1859 EVENT_LIST (6);
1860
1861 size_t cb = SvIV (cb_);
1862
1863 if (!SvOK (cb_))
1864 {
1865 NEED_SUCCESS (GetMemObjectInfo, (buf, CL_MEM_SIZE, sizeof (cb), &cb, 0));
1866 cb -= offset;
1867 }
1868
1869 cl_event ev;
1870 NEED_SUCCESS_ARG (void *ptr, EnqueueMapBuffer, (self, buf, blocking, map_flags, offset, cb, event_list_count, event_list_ptr, &ev, &res));
1871 XPUSHs (mapped_new (stash_mappedbuffer, self, buf, map_flags, ptr, cb, ev, 0, 0, 1, cb, 1, 1));
1872
1873void
1874map_image (OpenCL::Queue self, OpenCL::Image img, cl_bool blocking = 1, cl_map_flags map_flags = CL_MAP_READ | CL_MAP_WRITE, size_t x = 0, size_t y = 0, size_t z = 0, SV *width_ = &PL_sv_undef, SV *height_ = &PL_sv_undef, SV *depth_ = &PL_sv_undef, ...)
1875 ALIAS:
1876 enqueue_map_image = 0
1877 PPCODE:
1878 size_t width = SvIV (width_);
1879 if (!SvOK (width_))
1880 {
1881 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_WIDTH, sizeof (width), &width, 0));
1882 width -= x;
1883 }
1884
1885 size_t height = SvIV (width_);
1886 if (!SvOK (height_))
1887 {
1888 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_HEIGHT, sizeof (height), &height, 0));
1889 height -= y;
1890
1891 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1892 if (!height)
1893 height = 1;
1894 }
1895
1896 size_t depth = SvIV (width_);
1897 if (!SvOK (depth_))
1898 {
1899 NEED_SUCCESS (GetImageInfo, (img, CL_IMAGE_DEPTH, sizeof (depth), &depth, 0));
1900 depth -= z;
1901
1902 // stupid opencl returns 0 for depth, but requires 1 for 2d images
1903 if (!depth)
1904 depth = 1;
1905 }
1906
1907 const size_t origin[3] = { x, y, z };
1059 const size_t region[3] = { width, height, depth }; 1908 const size_t region[3] = { width, height, depth };
1909 size_t row_pitch, slice_pitch;
1060 EVENT_LIST (10, items - 10); 1910 EVENT_LIST (10);
1061 1911
1912 cl_event ev;
1913 NEED_SUCCESS_ARG (void *ptr, EnqueueMapImage, (self, img, blocking, map_flags, origin, region, &row_pitch, &slice_pitch, event_list_count, event_list_ptr, &ev, &res));
1914
1915 size_t cb = slice_pitch ? slice_pitch * region [2]
1916 : row_pitch ? row_pitch * region [1]
1917 : region [0];
1918
1919 XPUSHs (mapped_new (stash_mappedimage, self, img, map_flags, ptr, cb, ev, row_pitch, slice_pitch, 0, width, height, depth));
1920
1921void
1922unmap (OpenCL::Queue self, OpenCL::Mapped mapped, ...)
1923 PPCODE:
1924 mapped_unmap (cv, ST (1), mapped, self, &ST (2), items - 2);
1925 if (GIMME_V != G_VOID)
1926 {
1927 clRetainEvent (mapped->event);
1928 XPUSH_CLOBJ (stash_event, mapped->event);
1929 }
1930
1931void
1932task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1933 ALIAS:
1934 enqueue_task = 0
1935 PPCODE:
1936 EVENT_LIST (2);
1937
1938 cl_event ev = 0;
1062 NEED_SUCCESS (EnqueueCopyImageToBuffer, (self, src, dst, src_origin, region, dst_offset, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 1939 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1063 1940
1064 if (ev) 1941 if (ev)
1065 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 1942 XPUSH_CLOBJ (stash_event, ev);
1066 1943
1067void 1944void
1068enqueue_copy_buffer_to_image (OpenCL::Queue self, OpenCL::Buffer src, OpenCL::Image dst, size_t src_offset, size_t dst_x, size_t dst_y, size_t dst_z, size_t width, size_t height, size_t depth, ...)
1069 PPCODE:
1070 cl_event ev = 0;
1071 const size_t dst_origin[3] = { dst_x, dst_y, dst_z };
1072 const size_t region[3] = { width, height, depth };
1073 EVENT_LIST (10, items - 10);
1074
1075 NEED_SUCCESS (EnqueueCopyBufferToImage, (self, src, dst, src_offset, dst_origin, region, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1076
1077 if (ev)
1078 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1079
1080void
1081enqueue_task (OpenCL::Queue self, OpenCL::Kernel kernel, ...)
1082 PPCODE:
1083 cl_event ev = 0;
1084 EVENT_LIST (2, items - 2);
1085
1086 NEED_SUCCESS (EnqueueTask, (self, kernel, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1087
1088 if (ev)
1089 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1090
1091void
1092enqueue_nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...) 1945nd_range_kernel (OpenCL::Queue self, OpenCL::Kernel kernel, SV *global_work_offset, SV *global_work_size, SV *local_work_size = &PL_sv_undef, ...)
1946 ALIAS:
1947 enqueue_nd_range_kernel = 0
1093 PPCODE: 1948 PPCODE:
1094 cl_event ev = 0; 1949 EVENT_LIST (5);
1950
1095 size_t *gwo = 0, *gws, *lws = 0; 1951 size_t *gwo = 0, *gws, *lws = 0;
1096 int gws_len; 1952 int gws_len;
1097 size_t *lists; 1953 size_t *lists;
1098 int i; 1954 int i;
1099 EVENT_LIST (5, items - 5);
1100 1955
1101 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV) 1956 if (!SvROK (global_work_size) || SvTYPE (SvRV (global_work_size)) != SVt_PVAV)
1102 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference"); 1957 croak ("clEnqueueNDRangeKernel: global_work_size must be an array reference");
1103 1958
1104 gws_len = AvFILLp (SvRV (global_work_size)) + 1; 1959 gws_len = AvFILLp (SvRV (global_work_size)) + 1;
1105 1960
1106 lists = tmpbuf (sizeof (size_t) * 3 * gws_len); 1961 lists = tmpbuf (sizeof (size_t) * 3 * gws_len);
1107 1962
1108 gws = lists + gws_len * 0; 1963 gws = lists + gws_len * 0;
1109 for (i = 0; i < gws_len; ++i) 1964 for (i = 0; i < gws_len; ++i)
1965 {
1110 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]); 1966 gws [i] = SvIV (AvARRAY (SvRV (global_work_size))[i]);
1967 // at least nvidia crashes for 0-sized work group sizes, work around
1968 if (!gws [i])
1969 croak ("clEnqueueNDRangeKernel: global_work_size[%d] is zero, must be non-zero", i);
1970 }
1111 1971
1112 if (SvOK (global_work_offset)) 1972 if (SvOK (global_work_offset))
1113 { 1973 {
1114 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV) 1974 if (!SvROK (global_work_offset) || SvTYPE (SvRV (global_work_offset)) != SVt_PVAV)
1115 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference"); 1975 croak ("clEnqueueNDRangeKernel: global_work_offset must be undef or an array reference");
1123 } 1983 }
1124 1984
1125 if (SvOK (local_work_size)) 1985 if (SvOK (local_work_size))
1126 { 1986 {
1127 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV) 1987 if ((SvOK (local_work_size) && !SvROK (local_work_size)) || SvTYPE (SvRV (local_work_size)) != SVt_PVAV)
1128 croak ("clEnqueueNDRangeKernel: global_work_size must be undef or an array reference"); 1988 croak ("clEnqueueNDRangeKernel: local_work_size must be undef or an array reference");
1129 1989
1130 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len) 1990 if (AvFILLp (SvRV (local_work_size)) + 1 != gws_len)
1131 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size"); 1991 croak ("clEnqueueNDRangeKernel: local_work_local must be undef or an array of same size as global_work_size");
1132 1992
1133 lws = lists + gws_len * 2; 1993 lws = lists + gws_len * 2;
1134 for (i = 0; i < gws_len; ++i) 1994 for (i = 0; i < gws_len; ++i)
1995 {
1135 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]); 1996 lws [i] = SvIV (AvARRAY (SvRV (local_work_size))[i]);
1997 // at least nvidia crashes for 0-sized work group sizes, work around
1998 if (!lws [i])
1999 croak ("clEnqueueNDRangeKernel: local_work_size[%d] is zero, must be non-zero", i);
2000 }
1136 } 2001 }
1137 2002
2003 cl_event ev = 0;
1138 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2004 NEED_SUCCESS (EnqueueNDRangeKernel, (self, kernel, gws_len, gwo, gws, lws, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1139 2005
1140 if (ev) 2006 if (ev)
1141 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2007 XPUSH_CLOBJ (stash_event, ev);
2008
2009#if CL_VERSION_1_2
2010
2011void
2012migrate_mem_objects (OpenCL::Queue self, SV *objects, cl_mem_migration_flags flags, ...)
2013 ALIAS:
2014 enqueue_migrate_mem_objects = 0
2015 PPCODE:
2016 EVENT_LIST (3);
2017
2018 cl_uint obj_count;
2019 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
2020
2021 cl_event ev = 0;
2022 NEED_SUCCESS (EnqueueMigrateMemObjects, (self, obj_count, obj_list, flags, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2023
2024 if (ev)
2025 XPUSH_CLOBJ (stash_event, ev);
2026
2027#endif
1142 2028
1143#if cl_apple_gl_sharing || cl_khr_gl_sharing 2029#if cl_apple_gl_sharing || cl_khr_gl_sharing
1144 2030
1145void 2031void
1146enqueue_acquire_gl_objects (OpenCL::Queue self, SV *objects, ...) 2032acquire_gl_objects (OpenCL::Queue self, SV *objects, ...)
1147 ALIAS: 2033 ALIAS:
2034 release_gl_objects = 1
2035 enqueue_acquire_gl_objects = 0
1148 enqueue_release_gl_objects = 1 2036 enqueue_release_gl_objects = 1
1149 PPCODE: 2037 PPCODE:
1150 if (!SvROK (objects) || SvTYPE (SvRV (objects)) != SVt_PVAV) 2038 EVENT_LIST (2);
1151 croak ("OpenCL::Queue::enqueue_acquire/release_gl_objects argument 'objects' must be an arrayref with memory objects, in call"); 2039
2040 cl_uint obj_count;
2041 cl_mem *obj_list = object_list (cv, 0, "objects", objects, "OpenCL::Memory", &obj_count);
1152 2042
1153 cl_event ev = 0; 2043 cl_event ev = 0;
1154 EVENT_LIST (2, items - 2);
1155 AV *av = (AV *)SvRV (objects);
1156 cl_uint num_objects = av_len (av) + 1;
1157 cl_mem *object_list = tmpbuf (sizeof (cl_mem) * num_objects);
1158 int i;
1159
1160 for (i = num_objects; i--; )
1161 object_list [i] = SvPTROBJ ("OpenCL::Queue::enqueue_acquire/release_gl_objects", "objects", *av_fetch (av, i, 0), "OpenCL::Memory");
1162 2044
1163 if (ix) 2045 if (ix)
1164 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2046 NEED_SUCCESS (EnqueueReleaseGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1165 else 2047 else
1166 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, num_objects, object_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0)); 2048 NEED_SUCCESS (EnqueueAcquireGLObjects, (self, obj_count, obj_list, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
1167 2049
1168 if (ev) 2050 if (ev)
1169 XPUSH_NEW_OBJ ("OpenCL::Event", ev); 2051 XPUSH_CLOBJ (stash_event, ev);
1170 2052
1171#endif 2053#endif
1172 2054
1173#if !defined CL_VERSION_1_2 || defined CL_USE_DEPRECATED_OPENCL_1_1_APIS
1174
1175void 2055void
1176enqueue_marker (OpenCL::Queue self)
1177 PPCODE:
1178 cl_event ev;
1179 NEED_SUCCESS (EnqueueMarker, (self, &ev));
1180 XPUSH_NEW_OBJ ("OpenCL::Event", ev);
1181
1182void
1183enqueue_wait_for_events (OpenCL::Queue self, ...) 2056wait_for_events (OpenCL::Queue self, ...)
2057 ALIAS:
2058 enqueue_wait_for_events = 0
1184 CODE: 2059 CODE:
1185 EVENT_LIST (1, items - 1); 2060 EVENT_LIST (1);
2061#if PREFER_1_1
1186 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); 2062 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
1187 2063#else
1188void 2064 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, 0));
1189enqueue_barrier (OpenCL::Queue self)
1190 CODE:
1191 NEED_SUCCESS (EnqueueBarrier, (self));
1192
1193#endif 2065#endif
2066
2067void
2068marker (OpenCL::Queue self, ...)
2069 ALIAS:
2070 enqueue_marker = 0
2071 PPCODE:
2072 EVENT_LIST (1);
2073 cl_event ev = 0;
2074#if PREFER_1_1
2075 if (!event_list_count)
2076 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2077 else
2078#if CL_VERSION_1_2
2079 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2080#else
2081 {
2082 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr)); // also a barrier
2083 NEED_SUCCESS (EnqueueMarker, (self, GIMME_V != G_VOID ? &ev : 0));
2084 }
2085#endif
2086#else
2087 NEED_SUCCESS (EnqueueMarkerWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2088#endif
2089 if (ev)
2090 XPUSH_CLOBJ (stash_event, ev);
2091
2092void
2093barrier (OpenCL::Queue self, ...)
2094 ALIAS:
2095 enqueue_barrier = 0
2096 PPCODE:
2097 EVENT_LIST (1);
2098 cl_event ev = 0;
2099#if PREFER_1_1
2100 if (!event_list_count && GIMME_V == G_VOID)
2101 NEED_SUCCESS (EnqueueBarrier, (self));
2102 else
2103#if CL_VERSION_1_2
2104 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2105#else
2106 {
2107 if (event_list_count)
2108 NEED_SUCCESS (EnqueueWaitForEvents, (self, event_list_count, event_list_ptr));
2109
2110 if (GIMME_V != G_VOID)
2111 NEED_SUCCESS (EnqueueMarker, (self, &ev));
2112 }
2113#endif
2114#else
2115 NEED_SUCCESS (EnqueueBarrierWithWaitList, (self, event_list_count, event_list_ptr, GIMME_V != G_VOID ? &ev : 0));
2116#endif
2117 if (ev)
2118 XPUSH_CLOBJ (stash_event, ev);
1194 2119
1195void 2120void
1196flush (OpenCL::Queue self) 2121flush (OpenCL::Queue self)
1197 CODE: 2122 CODE:
1198 NEED_SUCCESS (Flush, (self)); 2123 NEED_SUCCESS (Flush, (self));
1214 PPCODE: 2139 PPCODE:
1215 cl_context value [1]; 2140 cl_context value [1];
1216 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0)); 2141 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_CONTEXT, sizeof (value), value, 0));
1217 EXTEND (SP, 1); 2142 EXTEND (SP, 1);
1218 const int i = 0; 2143 const int i = 0;
1219 {
1220 NEED_SUCCESS (RetainContext, (value [i])); 2144 NEED_SUCCESS (RetainContext, (value [i]));
1221 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2145 PUSH_CLOBJ (stash_context, value [i]);
1222 }
1223 2146
1224void 2147void
1225device (OpenCL::Queue self) 2148device (OpenCL::Queue self)
1226 PPCODE: 2149 PPCODE:
1227 cl_device_id value [1]; 2150 cl_device_id value [1];
1228 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0)); 2151 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_DEVICE, sizeof (value), value, 0));
1229 EXTEND (SP, 1); 2152 EXTEND (SP, 1);
1230 const int i = 0; 2153 const int i = 0;
1231 { 2154 PUSH_CLOBJ (stash_device, value [i]);
1232 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
1233 }
1234 2155
1235void 2156void
1236reference_count (OpenCL::Queue self) 2157reference_count (OpenCL::Queue self)
1237 PPCODE: 2158 PPCODE:
1238 cl_uint value [1]; 2159 cl_uint value [1];
1242 PUSHs (sv_2mortal (newSVuv (value [i]))); 2163 PUSHs (sv_2mortal (newSVuv (value [i])));
1243 2164
1244void 2165void
1245properties (OpenCL::Queue self) 2166properties (OpenCL::Queue self)
1246 PPCODE: 2167 PPCODE:
1247 cl_command_queue_properties value [1]; 2168 cl_ulong value [1];
1248 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0)); 2169 NEED_SUCCESS (GetCommandQueueInfo, (self, CL_QUEUE_PROPERTIES, sizeof (value), value, 0));
1249 EXTEND (SP, 1); 2170 EXTEND (SP, 1);
1250 const int i = 0; 2171 const int i = 0;
1251 PUSHs (sv_2mortal (newSViv (value [i]))); 2172 PUSHs (sv_2mortal (newSVuv (value [i])));
1252 2173
1253#END:command_queue 2174#END:command_queue
1254 2175
1255MODULE = OpenCL PACKAGE = OpenCL::Memory 2176MODULE = OpenCL PACKAGE = OpenCL::Memory
1256 2177
1262void 2183void
1263info (OpenCL::Memory self, cl_mem_info name) 2184info (OpenCL::Memory self, cl_mem_info name)
1264 PPCODE: 2185 PPCODE:
1265 INFO (MemObject) 2186 INFO (MemObject)
1266 2187
2188void
2189destructor_callback (OpenCL::Memory self, SV *notify)
2190 PPCODE:
2191 clSetMemObjectDestructorCallback (self, eq_destructor_notify, SvREFCNT_inc (s_get_cv (notify)));
2192
1267#BEGIN:mem 2193#BEGIN:mem
1268 2194
1269void 2195void
1270type (OpenCL::Memory self) 2196type (OpenCL::Memory self)
2197 ALIAS:
2198 type = CL_MEM_TYPE
2199 map_count = CL_MEM_MAP_COUNT
2200 reference_count = CL_MEM_REFERENCE_COUNT
1271 PPCODE: 2201 PPCODE:
1272 cl_mem_object_type value [1]; 2202 cl_uint value [1];
1273 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_TYPE, sizeof (value), value, 0)); 2203 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1274 EXTEND (SP, 1); 2204 EXTEND (SP, 1);
1275 const int i = 0; 2205 const int i = 0;
1276 PUSHs (sv_2mortal (newSViv (value [i]))); 2206 PUSHs (sv_2mortal (newSVuv (value [i])));
1277 2207
1278void 2208void
1279flags (OpenCL::Memory self) 2209flags (OpenCL::Memory self)
1280 PPCODE: 2210 PPCODE:
1281 cl_mem_flags value [1]; 2211 cl_ulong value [1];
1282 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0)); 2212 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_FLAGS, sizeof (value), value, 0));
1283 EXTEND (SP, 1); 2213 EXTEND (SP, 1);
1284 const int i = 0; 2214 const int i = 0;
1285 PUSHs (sv_2mortal (newSViv (value [i]))); 2215 PUSHs (sv_2mortal (newSVuv (value [i])));
1286 2216
1287void 2217void
1288size (OpenCL::Memory self) 2218size (OpenCL::Memory self)
1289 ALIAS: 2219 ALIAS:
1290 size = CL_MEM_SIZE 2220 size = CL_MEM_SIZE
1304 EXTEND (SP, 1); 2234 EXTEND (SP, 1);
1305 const int i = 0; 2235 const int i = 0;
1306 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i]))); 2236 PUSHs (sv_2mortal (newSVuv ((IV)(intptr_t)value [i])));
1307 2237
1308void 2238void
1309map_count (OpenCL::Memory self)
1310 ALIAS:
1311 map_count = CL_MEM_MAP_COUNT
1312 reference_count = CL_MEM_REFERENCE_COUNT
1313 PPCODE:
1314 cl_uint value [1];
1315 NEED_SUCCESS (GetMemObjectInfo, (self, ix, sizeof (value), value, 0));
1316 EXTEND (SP, 1);
1317 const int i = 0;
1318 PUSHs (sv_2mortal (newSVuv (value [i])));
1319
1320void
1321context (OpenCL::Memory self) 2239context (OpenCL::Memory self)
1322 PPCODE: 2240 PPCODE:
1323 cl_context value [1]; 2241 cl_context value [1];
1324 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0)); 2242 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_CONTEXT, sizeof (value), value, 0));
1325 EXTEND (SP, 1); 2243 EXTEND (SP, 1);
1326 const int i = 0; 2244 const int i = 0;
1327 {
1328 NEED_SUCCESS (RetainContext, (value [i])); 2245 NEED_SUCCESS (RetainContext, (value [i]));
1329 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2246 PUSH_CLOBJ (stash_context, value [i]);
1330 }
1331 2247
1332void 2248void
1333associated_memobject (OpenCL::Memory self) 2249associated_memobject (OpenCL::Memory self)
1334 PPCODE: 2250 PPCODE:
1335 cl_mem value [1]; 2251 cl_mem value [1];
1336 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0)); 2252 NEED_SUCCESS (GetMemObjectInfo, (self, CL_MEM_ASSOCIATED_MEMOBJECT, sizeof (value), value, 0));
1337 EXTEND (SP, 1); 2253 EXTEND (SP, 1);
1338 const int i = 0; 2254 const int i = 0;
1339 {
1340 NEED_SUCCESS (RetainMemObject, (value [i])); 2255 NEED_SUCCESS (RetainMemObject, (value [i]));
1341 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Memory", value [i])); 2256 PUSH_CLOBJ (stash_memory, value [i]);
1342 }
1343 2257
1344#END:mem 2258#END:mem
1345 2259
1346#if cl_apple_gl_sharing || cl_khr_gl_sharing 2260#if cl_apple_gl_sharing || cl_khr_gl_sharing
1347 2261
1366 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!"); 2280 croak ("clCreateSubBuffer: cannot use/copy/alloc host ptr, doesn't make sense, check your flags!");
1367 2281
1368 cl_buffer_region crdata = { origin, size }; 2282 cl_buffer_region crdata = { origin, size };
1369 2283
1370 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res)); 2284 NEED_SUCCESS_ARG (cl_mem mem, CreateSubBuffer, (self, flags, CL_BUFFER_CREATE_TYPE_REGION, &crdata, &res));
1371 XPUSH_NEW_OBJ ("OpenCL::Buffer", mem); 2285 XPUSH_CLOBJ (stash_buffer, mem);
1372 2286
1373MODULE = OpenCL PACKAGE = OpenCL::Image 2287MODULE = OpenCL PACKAGE = OpenCL::Image
1374 2288
1375void 2289void
1376image_info (OpenCL::Image self, cl_image_info name) 2290image_info (OpenCL::Image self, cl_image_info name)
1377 PPCODE: 2291 PPCODE:
1378 INFO (Image) 2292 INFO (Image)
2293
2294void
2295format (OpenCL::Image self)
2296 PPCODE:
2297 cl_image_format format;
2298 NEED_SUCCESS (GetImageInfo, (self, CL_IMAGE_FORMAT, sizeof (format), &format, 0));
2299 EXTEND (SP, 2);
2300 PUSHs (sv_2mortal (newSVuv (format.image_channel_order)));
2301 PUSHs (sv_2mortal (newSVuv (format.image_channel_data_type)));
1379 2302
1380#BEGIN:image 2303#BEGIN:image
1381 2304
1382void 2305void
1383element_size (OpenCL::Image self) 2306element_size (OpenCL::Image self)
1437 2360
1438#BEGIN:sampler 2361#BEGIN:sampler
1439 2362
1440void 2363void
1441reference_count (OpenCL::Sampler self) 2364reference_count (OpenCL::Sampler self)
2365 ALIAS:
2366 reference_count = CL_SAMPLER_REFERENCE_COUNT
2367 normalized_coords = CL_SAMPLER_NORMALIZED_COORDS
2368 addressing_mode = CL_SAMPLER_ADDRESSING_MODE
1442 PPCODE: 2369 PPCODE:
1443 cl_uint value [1]; 2370 cl_uint value [1];
1444 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_REFERENCE_COUNT, sizeof (value), value, 0)); 2371 NEED_SUCCESS (GetSamplerInfo, (self, ix, sizeof (value), value, 0));
1445 EXTEND (SP, 1); 2372 EXTEND (SP, 1);
1446 const int i = 0; 2373 const int i = 0;
1447 PUSHs (sv_2mortal (newSVuv (value [i]))); 2374 PUSHs (sv_2mortal (newSVuv (value [i])));
1448 2375
1449void 2376void
1451 PPCODE: 2378 PPCODE:
1452 cl_context value [1]; 2379 cl_context value [1];
1453 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0)); 2380 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_CONTEXT, sizeof (value), value, 0));
1454 EXTEND (SP, 1); 2381 EXTEND (SP, 1);
1455 const int i = 0; 2382 const int i = 0;
1456 {
1457 NEED_SUCCESS (RetainContext, (value [i])); 2383 NEED_SUCCESS (RetainContext, (value [i]));
1458 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2384 PUSH_CLOBJ (stash_context, value [i]);
1459 }
1460
1461void
1462normalized_coords (OpenCL::Sampler self)
1463 PPCODE:
1464 cl_addressing_mode value [1];
1465 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_NORMALIZED_COORDS, sizeof (value), value, 0));
1466 EXTEND (SP, 1);
1467 const int i = 0;
1468 PUSHs (sv_2mortal (newSViv (value [i])));
1469
1470void
1471addressing_mode (OpenCL::Sampler self)
1472 PPCODE:
1473 cl_filter_mode value [1];
1474 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_ADDRESSING_MODE, sizeof (value), value, 0));
1475 EXTEND (SP, 1);
1476 const int i = 0;
1477 PUSHs (sv_2mortal (newSViv (value [i])));
1478 2385
1479void 2386void
1480filter_mode (OpenCL::Sampler self) 2387filter_mode (OpenCL::Sampler self)
1481 PPCODE: 2388 PPCODE:
1482 cl_bool value [1]; 2389 cl_uint value [1];
1483 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0)); 2390 NEED_SUCCESS (GetSamplerInfo, (self, CL_SAMPLER_FILTER_MODE, sizeof (value), value, 0));
1484 EXTEND (SP, 1); 2391 EXTEND (SP, 1);
1485 const int i = 0; 2392 const int i = 0;
1486 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no)); 2393 PUSHs (sv_2mortal (value [i] ? &PL_sv_yes : &PL_sv_no));
1487 2394
1493DESTROY (OpenCL::Program self) 2400DESTROY (OpenCL::Program self)
1494 CODE: 2401 CODE:
1495 clReleaseProgram (self); 2402 clReleaseProgram (self);
1496 2403
1497void 2404void
1498build (OpenCL::Program self, OpenCL::Device device, SV *options = &PL_sv_undef) 2405build (OpenCL::Program self, SV *devices = &PL_sv_undef, SV *options = &PL_sv_undef, SV *notify = &PL_sv_undef)
2406 ALIAS:
2407 build_async = 1
1499 CODE: 2408 CODE:
2409 cl_uint device_count = 0;
2410 cl_device_id *device_list = 0;
2411
2412 if (SvOK (devices))
2413 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2414
2415 void *user_data;
2416 program_callback pfn_notify = make_program_callback (notify, &user_data);
2417
2418 if (ix)
2419 build_program_async (self, device_count, device_list, SvPVbyte_nolen (options), user_data);
2420 else
2421 NEED_SUCCESS (BuildProgram, (self, device_count, device_list, SvPVbyte_nolen (options), pfn_notify, user_data));
2422
2423#if CL_VERSION_1_2
2424
2425void
2426compile (OpenCL::Program self, SV *devices, SV *options = &PL_sv_undef, SV *headers = &PL_sv_undef, SV *notify = &PL_sv_undef)
2427 CODE:
2428 cl_uint device_count = 0;
2429 cl_device_id *device_list = 0;
2430
2431 if (SvOK (devices))
2432 device_list = object_list (cv, 1, "devices", devices, "OpenCL::Device", &device_count);
2433
2434 cl_uint header_count = 0;
2435 cl_program *header_list = 0;
2436 const char **header_name = 0;
2437
2438 if (SvOK (headers))
2439 {
2440 if (!SvROK (devices) || SvTYPE (SvRV (devices)) != SVt_PVHV)
2441 croak ("clCompileProgram: headers must be undef or a hashref of name => OpenCL::Program pairs");
2442
2443 HV *hv = (HV *)SvRV (devices);
2444
2445 header_count = hv_iterinit (hv);
2446 header_list = tmpbuf (sizeof (*header_list) * header_count);
2447 header_name = tmpbuf (sizeof (*header_name) * header_count);
2448
2449 HE *he;
2450 int i = 0;
2451 while (he = hv_iternext (hv))
2452 {
2453 header_name [i] = SvPVbyte_nolen (HeSVKEY_force (he));
2454 header_list [i] = SvCLOBJ (cv, "headers", HeVAL (he), "OpenCL::Program");
2455 ++i;
2456 }
2457 }
2458
2459 void *user_data;
2460 program_callback pfn_notify = make_program_callback (notify, &user_data);
2461
1500 NEED_SUCCESS (BuildProgram, (self, 1, &device, SvPVbyte_nolen (options), 0, 0)); 2462 NEED_SUCCESS (CompileProgram, (self, device_count, device_list, SvPVbyte_nolen (options),
2463 header_count, header_list, header_name, pfn_notify, user_data));
2464
2465#endif
1501 2466
1502void 2467void
1503build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name) 2468build_info (OpenCL::Program self, OpenCL::Device device, cl_program_build_info name)
1504 PPCODE: 2469 PPCODE:
1505 size_t size; 2470 size_t size;
1514#BEGIN:program_build 2479#BEGIN:program_build
1515 2480
1516void 2481void
1517build_status (OpenCL::Program self, OpenCL::Device device) 2482build_status (OpenCL::Program self, OpenCL::Device device)
1518 PPCODE: 2483 PPCODE:
1519 cl_build_status value [1]; 2484 cl_int value [1];
1520 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0)); 2485 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BUILD_STATUS, sizeof (value), value, 0));
1521 EXTEND (SP, 1); 2486 EXTEND (SP, 1);
1522 const int i = 0; 2487 const int i = 0;
1523 PUSHs (sv_2mortal (newSViv (value [i]))); 2488 PUSHs (sv_2mortal (newSViv (value [i])));
1524 2489
1534 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0)); 2499 NEED_SUCCESS (GetProgramBuildInfo, (self, device, ix, size, value, 0));
1535 EXTEND (SP, 1); 2500 EXTEND (SP, 1);
1536 const int i = 0; 2501 const int i = 0;
1537 PUSHs (sv_2mortal (newSVpv (value, 0))); 2502 PUSHs (sv_2mortal (newSVpv (value, 0)));
1538 2503
2504void
2505binary_type (OpenCL::Program self, OpenCL::Device device)
2506 PPCODE:
2507 cl_uint value [1];
2508 NEED_SUCCESS (GetProgramBuildInfo, (self, device, CL_PROGRAM_BINARY_TYPE, sizeof (value), value, 0));
2509 EXTEND (SP, 1);
2510 const int i = 0;
2511 PUSHs (sv_2mortal (newSVuv ((UV)value [i])));
2512
1539#END:program_build 2513#END:program_build
1540 2514
1541void 2515void
1542kernel (OpenCL::Program program, SV *function) 2516kernel (OpenCL::Program program, SV *function)
1543 PPCODE: 2517 PPCODE:
1544 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res)); 2518 NEED_SUCCESS_ARG (cl_kernel kernel, CreateKernel, (program, SvPVbyte_nolen (function), &res));
1545 XPUSH_NEW_OBJ ("OpenCL::Kernel", kernel); 2519 XPUSH_CLOBJ (stash_kernel, kernel);
2520
2521void
2522kernels_in_program (OpenCL::Program program)
2523 PPCODE:
2524 cl_uint num_kernels;
2525 NEED_SUCCESS (CreateKernelsInProgram, (program, 0, 0, &num_kernels));
2526 cl_kernel *kernels = tmpbuf (sizeof (cl_kernel) * num_kernels);
2527 NEED_SUCCESS (CreateKernelsInProgram, (program, num_kernels, kernels, 0));
2528
2529 int i;
2530 EXTEND (SP, num_kernels);
2531 for (i = 0; i < num_kernels; ++i)
2532 PUSH_CLOBJ (stash_kernel, kernels [i]);
1546 2533
1547void 2534void
1548info (OpenCL::Program self, cl_program_info name) 2535info (OpenCL::Program self, cl_program_info name)
1549 PPCODE: 2536 PPCODE:
1550 INFO (Program) 2537 INFO (Program)
1596 PPCODE: 2583 PPCODE:
1597 cl_context value [1]; 2584 cl_context value [1];
1598 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0)); 2585 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_CONTEXT, sizeof (value), value, 0));
1599 EXTEND (SP, 1); 2586 EXTEND (SP, 1);
1600 const int i = 0; 2587 const int i = 0;
1601 {
1602 NEED_SUCCESS (RetainContext, (value [i])); 2588 NEED_SUCCESS (RetainContext, (value [i]));
1603 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2589 PUSH_CLOBJ (stash_context, value [i]);
1604 }
1605 2590
1606void 2591void
1607devices (OpenCL::Program self) 2592devices (OpenCL::Program self)
1608 PPCODE: 2593 PPCODE:
1609 size_t size; 2594 size_t size;
1611 cl_device_id *value = tmpbuf (size); 2596 cl_device_id *value = tmpbuf (size);
1612 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0)); 2597 NEED_SUCCESS (GetProgramInfo, (self, CL_PROGRAM_DEVICES, size, value, 0));
1613 int i, n = size / sizeof (*value); 2598 int i, n = size / sizeof (*value);
1614 EXTEND (SP, n); 2599 EXTEND (SP, n);
1615 for (i = 0; i < n; ++i) 2600 for (i = 0; i < n; ++i)
1616 { 2601 PUSH_CLOBJ (stash_device, value [i]);
1617 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Device", value [i]));
1618 }
1619 2602
1620void 2603void
1621source (OpenCL::Program self) 2604source (OpenCL::Program self)
1622 PPCODE: 2605 PPCODE:
1623 size_t size; 2606 size_t size;
1648DESTROY (OpenCL::Kernel self) 2631DESTROY (OpenCL::Kernel self)
1649 CODE: 2632 CODE:
1650 clReleaseKernel (self); 2633 clReleaseKernel (self);
1651 2634
1652void 2635void
2636setf (OpenCL::Kernel self, const char *format, ...)
2637 CODE:
2638 int i;
2639 for (i = 2; ; ++i)
2640 {
2641 while (*format == ' ')
2642 ++format;
2643
2644 char type = *format++;
2645
2646 if (!type)
2647 break;
2648
2649 if (i >= items)
2650 croak ("OpenCL::Kernel::setf format string too long (not enough arguments)");
2651
2652 SV *sv = ST (i);
2653
2654 union
2655 {
2656 cl_char cc; cl_uchar cC; cl_short cs; cl_ushort cS;
2657 cl_int ci; cl_uint cI; cl_long cl; cl_ulong cL;
2658 cl_half ch; cl_float cf; cl_double cd;
2659 cl_mem cm;
2660 cl_sampler ca;
2661 size_t cz;
2662 cl_event ce;
2663 } arg;
2664 size_t size;
2665 int nullarg = 0;
2666
2667 switch (type)
2668 {
2669 case 'c': arg.cc = SvIV (sv); size = sizeof (arg.cc); break;
2670 case 'C': arg.cC = SvUV (sv); size = sizeof (arg.cC); break;
2671 case 's': arg.cs = SvIV (sv); size = sizeof (arg.cs); break;
2672 case 'S': arg.cS = SvUV (sv); size = sizeof (arg.cS); break;
2673 case 'i': arg.ci = SvIV (sv); size = sizeof (arg.ci); break;
2674 case 'I': arg.cI = SvUV (sv); size = sizeof (arg.cI); break;
2675 case 'l': arg.cl = SvIV (sv); size = sizeof (arg.cl); break;
2676 case 'L': arg.cL = SvUV (sv); size = sizeof (arg.cL); break;
2677
2678 case 'h': arg.ch = SvUV (sv); size = sizeof (arg.ch); break;
2679 case 'f': arg.cf = SvNV (sv); size = sizeof (arg.cf); break;
2680 case 'd': arg.cd = SvNV (sv); size = sizeof (arg.cd); break;
2681
2682 case 'z': nullarg = 1; size = SvIV (sv); break;
2683
2684 case 'm': nullarg = !SvOK (sv); arg.cm = SvCLOBJ (cv, "m", sv, "OpenCL::Memory" ); size = sizeof (arg.cm); break;
2685 case 'a': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "a", sv, "OpenCL::Sampler"); size = sizeof (arg.ca); break;
2686 case 'e': nullarg = !SvOK (sv); arg.ca = SvCLOBJ (cv, "e", sv, "OpenCL::Event" ); size = sizeof (arg.ce); break;
2687
2688 default:
2689 croak ("OpenCL::Kernel::setf format character '%c' not supported", type);
2690 }
2691
2692 res = clSetKernelArg (self, i - 2, size, nullarg ? 0 : &arg);
2693 if (res)
2694 croak ("OpenCL::Kernel::setf kernel parameter '%c' (#%d): %s", type, i - 2, err2str (res));
2695 }
2696
2697 if (i != items)
2698 croak ("OpenCL::Kernel::setf format string too short (too many arguments)");
2699
2700void
1653set_char (OpenCL::Kernel self, cl_uint idx, cl_char value) 2701set_char (OpenCL::Kernel self, cl_uint idx, cl_char value)
1654 CODE: 2702 CODE:
1655 clSetKernelArg (self, idx, sizeof (value), &value); 2703 clSetKernelArg (self, idx, sizeof (value), &value);
1656 2704
1657void 2705void
1705 clSetKernelArg (self, idx, sizeof (value), &value); 2753 clSetKernelArg (self, idx, sizeof (value), &value);
1706 2754
1707void 2755void
1708set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value) 2756set_memory (OpenCL::Kernel self, cl_uint idx, OpenCL::Memory_ornull value)
1709 CODE: 2757 CODE:
1710 clSetKernelArg (self, idx, sizeof (value), &value); 2758 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1711 2759
1712void 2760void
1713set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value) 2761set_buffer (OpenCL::Kernel self, cl_uint idx, OpenCL::Buffer_ornull value)
1714 CODE: 2762 CODE:
1715 clSetKernelArg (self, idx, sizeof (value), &value); 2763 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1716 2764
1717void 2765void
1718set_image2d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image2D_ornull value) 2766set_image (OpenCL::Kernel self, cl_uint idx, OpenCL::Image_ornull value)
1719 CODE: 2767 CODE:
1720 clSetKernelArg (self, idx, sizeof (value), &value); 2768 clSetKernelArg (self, idx, sizeof (value), value ? &value : 0);
1721
1722void
1723set_image3d (OpenCL::Kernel self, cl_uint idx, OpenCL::Image3D_ornull value)
1724 CODE:
1725 clSetKernelArg (self, idx, sizeof (value), &value);
1726 2769
1727void 2770void
1728set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value) 2771set_sampler (OpenCL::Kernel self, cl_uint idx, OpenCL::Sampler value)
1729 CODE: 2772 CODE:
1730 clSetKernelArg (self, idx, sizeof (value), &value); 2773 clSetKernelArg (self, idx, sizeof (value), &value);
1774 PPCODE: 2817 PPCODE:
1775 cl_context value [1]; 2818 cl_context value [1];
1776 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0)); 2819 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_CONTEXT, sizeof (value), value, 0));
1777 EXTEND (SP, 1); 2820 EXTEND (SP, 1);
1778 const int i = 0; 2821 const int i = 0;
1779 {
1780 NEED_SUCCESS (RetainContext, (value [i])); 2822 NEED_SUCCESS (RetainContext, (value [i]));
1781 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2823 PUSH_CLOBJ (stash_context, value [i]);
1782 }
1783 2824
1784void 2825void
1785program (OpenCL::Kernel self) 2826program (OpenCL::Kernel self)
1786 PPCODE: 2827 PPCODE:
1787 cl_program value [1]; 2828 cl_program value [1];
1788 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0)); 2829 NEED_SUCCESS (GetKernelInfo, (self, CL_KERNEL_PROGRAM, sizeof (value), value, 0));
1789 EXTEND (SP, 1); 2830 EXTEND (SP, 1);
1790 const int i = 0; 2831 const int i = 0;
1791 {
1792 NEED_SUCCESS (RetainProgram, (value [i])); 2832 NEED_SUCCESS (RetainProgram, (value [i]));
1793 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Program", value [i])); 2833 PUSH_CLOBJ (stash_program, value [i]);
1794 }
1795 2834
1796#END:kernel 2835#END:kernel
1797 2836
1798void 2837void
1799work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name) 2838work_group_info (OpenCL::Kernel self, OpenCL::Device device, cl_kernel_work_group_info name)
1845 const int i = 0; 2884 const int i = 0;
1846 PUSHs (sv_2mortal (newSVuv (value [i]))); 2885 PUSHs (sv_2mortal (newSVuv (value [i])));
1847 2886
1848#END:kernel_work_group 2887#END:kernel_work_group
1849 2888
2889#if CL_VERSION_1_2
2890
2891void
2892arg_info (OpenCL::Kernel self, cl_uint idx, cl_kernel_arg_info name)
2893 PPCODE:
2894 size_t size;
2895 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, 0, 0, &size));
2896 SV *sv = sv_2mortal (newSV (size));
2897 SvUPGRADE (sv, SVt_PV);
2898 SvPOK_only (sv);
2899 SvCUR_set (sv, size);
2900 NEED_SUCCESS (GetKernelArgInfo, (self, idx, name, size, SvPVX (sv), 0));
2901 XPUSHs (sv);
2902
2903#BEGIN:kernel_arg
2904
2905void
2906arg_address_qualifier (OpenCL::Kernel self, cl_uint idx)
2907 ALIAS:
2908 arg_address_qualifier = CL_KERNEL_ARG_ADDRESS_QUALIFIER
2909 arg_access_qualifier = CL_KERNEL_ARG_ACCESS_QUALIFIER
2910 PPCODE:
2911 cl_uint value [1];
2912 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, sizeof (value), value, 0));
2913 EXTEND (SP, 1);
2914 const int i = 0;
2915 PUSHs (sv_2mortal (newSVuv (value [i])));
2916
2917void
2918arg_type_name (OpenCL::Kernel self, cl_uint idx)
2919 ALIAS:
2920 arg_type_name = CL_KERNEL_ARG_TYPE_NAME
2921 arg_name = CL_KERNEL_ARG_NAME
2922 PPCODE:
2923 size_t size;
2924 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, 0, 0, &size));
2925 char *value = tmpbuf (size);
2926 NEED_SUCCESS (GetKernelArgInfo, (self, idx, ix, size, value, 0));
2927 EXTEND (SP, 1);
2928 const int i = 0;
2929 PUSHs (sv_2mortal (newSVpv (value, 0)));
2930
2931void
2932arg_type_qualifier (OpenCL::Kernel self, cl_uint idx)
2933 PPCODE:
2934 cl_ulong value [1];
2935 NEED_SUCCESS (GetKernelArgInfo, (self, idx, CL_KERNEL_ARG_TYPE_QUALIFIER, sizeof (value), value, 0));
2936 EXTEND (SP, 1);
2937 const int i = 0;
2938 PUSHs (sv_2mortal (newSVuv (value [i])));
2939
2940#END:kernel_arg
2941
2942#endif
2943
1850MODULE = OpenCL PACKAGE = OpenCL::Event 2944MODULE = OpenCL PACKAGE = OpenCL::Event
1851 2945
1852void 2946void
1853DESTROY (OpenCL::Event self) 2947DESTROY (OpenCL::Event self)
1854 CODE: 2948 CODE:
1856 2950
1857void 2951void
1858wait (OpenCL::Event self) 2952wait (OpenCL::Event self)
1859 CODE: 2953 CODE:
1860 clWaitForEvents (1, &self); 2954 clWaitForEvents (1, &self);
2955
2956void
2957cb (OpenCL::Event self, cl_int command_exec_callback_type, SV *cb)
2958 CODE:
2959 clSetEventCallback (self, command_exec_callback_type, eq_event_notify, SvREFCNT_inc (s_get_cv (cb)));
1861 2960
1862void 2961void
1863info (OpenCL::Event self, cl_event_info name) 2962info (OpenCL::Event self, cl_event_info name)
1864 PPCODE: 2963 PPCODE:
1865 INFO (Event) 2964 INFO (Event)
1871 PPCODE: 2970 PPCODE:
1872 cl_command_queue value [1]; 2971 cl_command_queue value [1];
1873 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0)); 2972 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_QUEUE, sizeof (value), value, 0));
1874 EXTEND (SP, 1); 2973 EXTEND (SP, 1);
1875 const int i = 0; 2974 const int i = 0;
1876 {
1877 NEED_SUCCESS (RetainCommandQueue, (value [i])); 2975 NEED_SUCCESS (RetainCommandQueue, (value [i]));
1878 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Queue", value [i])); 2976 PUSH_CLOBJ (stash_queue, value [i]);
1879 }
1880 2977
1881void 2978void
1882command_type (OpenCL::Event self) 2979command_type (OpenCL::Event self)
1883 PPCODE:
1884 cl_command_type value [1];
1885 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_COMMAND_TYPE, sizeof (value), value, 0));
1886 EXTEND (SP, 1);
1887 const int i = 0;
1888 PUSHs (sv_2mortal (newSVuv (value [i])));
1889
1890void
1891reference_count (OpenCL::Event self)
1892 ALIAS: 2980 ALIAS:
2981 command_type = CL_EVENT_COMMAND_TYPE
1893 reference_count = CL_EVENT_REFERENCE_COUNT 2982 reference_count = CL_EVENT_REFERENCE_COUNT
1894 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS 2983 command_execution_status = CL_EVENT_COMMAND_EXECUTION_STATUS
1895 PPCODE: 2984 PPCODE:
1896 cl_uint value [1]; 2985 cl_uint value [1];
1897 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0)); 2986 NEED_SUCCESS (GetEventInfo, (self, ix, sizeof (value), value, 0));
1904 PPCODE: 2993 PPCODE:
1905 cl_context value [1]; 2994 cl_context value [1];
1906 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0)); 2995 NEED_SUCCESS (GetEventInfo, (self, CL_EVENT_CONTEXT, sizeof (value), value, 0));
1907 EXTEND (SP, 1); 2996 EXTEND (SP, 1);
1908 const int i = 0; 2997 const int i = 0;
1909 {
1910 NEED_SUCCESS (RetainContext, (value [i])); 2998 NEED_SUCCESS (RetainContext, (value [i]));
1911 PUSHs (NEW_MORTAL_OBJ ("OpenCL::Context", value [i])); 2999 PUSH_CLOBJ (stash_context, value [i]);
1912 }
1913 3000
1914#END:event 3001#END:event
1915 3002
1916void 3003void
1917profiling_info (OpenCL::Event self, cl_profiling_info name) 3004profiling_info (OpenCL::Event self, cl_profiling_info name)
1941void 3028void
1942set_status (OpenCL::UserEvent self, cl_int execution_status) 3029set_status (OpenCL::UserEvent self, cl_int execution_status)
1943 CODE: 3030 CODE:
1944 clSetUserEventStatus (self, execution_status); 3031 clSetUserEventStatus (self, execution_status);
1945 3032
3033MODULE = OpenCL PACKAGE = OpenCL::Mapped
3034
3035void
3036DESTROY (SV *self)
3037 CODE:
3038 OpenCL__Mapped mapped = SvMAPPED (self);
3039
3040 clEnqueueUnmapMemObject (mapped->queue, mapped->memobj, mapped->ptr, 1, &mapped->event, 0);
3041 mapped_detach (self, mapped);
3042
3043 clReleaseCommandQueue (mapped->queue);
3044 clReleaseEvent (mapped->event);
3045 Safefree (mapped);
3046
3047void
3048unmap (OpenCL::Mapped self, ...)
3049 CODE:
3050 mapped_unmap (cv, ST (0), self, self->queue, &ST (1), items - 1);
3051
3052bool
3053mapped (OpenCL::Mapped self)
3054 CODE:
3055 RETVAL = !!self->ptr;
3056 OUTPUT:
3057 RETVAL
3058
3059void
3060wait (OpenCL::Mapped self)
3061 PPCODE:
3062 if (self->event)
3063 NEED_SUCCESS (WaitForEvents, (1, &self->event));
3064
3065void
3066event (OpenCL::Mapped self)
3067 PPCODE:
3068 if (!self->event)
3069 XSRETURN_UNDEF;
3070
3071 clRetainEvent (self->event);
3072 XPUSH_CLOBJ (stash_event, self->event);
3073
3074#define MAPPED_OFFSET_CB offsetof (struct mapped, cb)
3075#define MAPPED_OFFSET_ROW_PITCH offsetof (struct mapped, row_pitch)
3076#define MAPPED_OFFSET_SLICE_PITCH offsetof (struct mapped, slice_pitch)
3077#define MAPPED_OFFSET_WIDTH offsetof (struct mapped, width)
3078#define MAPPED_OFFSET_HEIGHT offsetof (struct mapped, height)
3079#define MAPPED_OFFSET_DEPTH offsetof (struct mapped, depth)
3080
3081IV
3082size (OpenCL::Mapped self)
3083 ALIAS:
3084 size = MAPPED_OFFSET_CB
3085 row_pitch = MAPPED_OFFSET_ROW_PITCH
3086 slice_pitch = MAPPED_OFFSET_SLICE_PITCH
3087 width = MAPPED_OFFSET_WIDTH
3088 height = MAPPED_OFFSET_HEIGHT
3089 depth = MAPPED_OFFSET_DEPTH
3090 CODE:
3091 RETVAL = *(size_t *)((char *)self + ix);
3092 OUTPUT:
3093 RETVAL
3094
3095IV
3096ptr (OpenCL::Mapped self)
3097 CODE:
3098 RETVAL = PTR2IV (self->ptr);
3099 OUTPUT:
3100 RETVAL
3101
3102void
3103set (OpenCL::Mapped self, size_t offset, SV *data)
3104 CODE:
3105 STRLEN len;
3106 const char *ptr = SvPVbyte (data, len);
3107
3108 if (offset + len > self->cb)
3109 croak ("OpenCL::Mapped::set out of bound condition detected");
3110
3111 memcpy (offset + (char *)self->ptr, ptr, len);
3112
3113void
3114get_row (OpenCL::Mapped self, size_t count, size_t x = 0, size_t y = 0, size_t z = 0)
3115 PPCODE:
3116 if (!SvOK (ST (1)))
3117 count = self->width - x;
3118
3119 if (x + count > self->width)
3120 croak ("OpenCL::Mapped::get: x + count crosses a row boundary");
3121
3122 if (y >= self->height)
3123 croak ("OpenCL::Mapped::get: y coordinate out of bounds");
3124
3125 if (z >= self->depth)
3126 croak ("OpenCL::Mapped::get: z coordinate out of bounds");
3127
3128 size_t element = mapped_element_size (self);
3129
3130 count *= element;
3131 x *= element;
3132
3133 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3134 XPUSHs (sv_2mortal (newSVpvn (ptr, count)));
3135
3136void
3137set_row (OpenCL::Mapped self, SV *data, size_t x = 0, size_t y = 0, size_t z = 0)
3138 PPCODE:
3139 STRLEN count;
3140 char *dataptr = SvPVbyte (data, count);
3141 size_t element = mapped_element_size (self);
3142
3143 x *= element;
3144
3145 if (x + count > self->width * element)
3146 croak ("OpenCL::Mapped::set: x + data size crosses a row boundary");
3147
3148 if (y >= self->height)
3149 croak ("OpenCL::Mapped::set: y coordinate out of bounds");
3150
3151 if (z >= self->depth)
3152 croak ("OpenCL::Mapped::set: z coordinate out of bounds");
3153
3154 char *ptr = (char *)self->ptr + x + y * self->row_pitch + z * self->slice_pitch;
3155 memcpy (ptr, dataptr, count);
3156
3157MODULE = OpenCL PACKAGE = OpenCL::MappedBuffer
3158
3159MODULE = OpenCL PACKAGE = OpenCL::MappedImage
3160
3161IV
3162element_size (OpenCL::Mapped self)
3163 CODE:
3164 RETVAL = mapped_element_size (self);
3165 OUTPUT:
3166 RETVAL
3167

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines